Explainable artificial intelligence approach towards classifying educational android app reviews using deep learning

Mobile application developers rely largely on user reviews for identifying issues in mobile applications and meeting the users' expectations. User reviews are unstructured, unorganized and very informal. Identifying and classifying issues by extracting required information from reviews is difficult due to a large number of reviews. To automate the process of classifying reviews many researchers have adopted machine learning approaches. Keeping in view, the rising demand for educational applications, especially during COVID-19, this research aims to automate Android application education reviews' classification and sentiment analysis using natural language processing and machine learning techniques. A baseline corpus comprising 13,000 records has been built by collecting reviews of more than 20 educational applications. The reviews were then manually labelled with respect to sentiment and issue types mentioned in each review. User reviews are classified into eight categories and various machine learning algorithms are applied to classify users' sentiments and issues of applications. The results demonstrate that our proposed framework achieved an accuracy of 97% for sentiment identification and an accuracy of 94% in classifying the most significant issues. Moreover, the interpretability of the model is verified by using the explainable artificial intelligence technique of local interpretable model-agnostic explanations. (PsycInfo Database Record (c) 2023 APA, all rights reserved)

Pediatric Cardiac Screening Data Warehouse: A National Public Health Resource

Background: Cardiac screening of youth for prevention of sudden cardiac death in the young (SCDY) has been debated due to the absence of large population-specific screening data with outcomes. Despite years of screening by US public screening groups (PSG), there is minimal coordination of effort and no standardized methods for real-world data collection. Objective(s): To understand the methods, quality, outcomes, and best practices of youth screening, the Cardiac Safety Research Consortium Pediatric Cardiology Working Group, in collaboration with FDA and PSGs, developed and enabled a scalable system to collect a uniform pediatric cardiac screening dataset including digital ECGs and post-screening electronic follow-up data. Method(s): Front end data collection (figure) was developed to include use of a universal unique ID system to align paper/digital collection of health and ECG data. PSGs use secure data transfer portals for digital ECG data upload for conversion to device-agnostic standardized FDA format to store in the national pediatric cardiac screening data warehouse. Follow-up data are obtained at designated post-screening intervals (one week, one and 3 months for pilot study) using initial text message contact followed by electronic consent (REDCap) and answering online health surveys. Result(s): Fourteen PSGs in ten states participated in the pilot study. PSG warehouse data include 33840 retrospective ECG datasets collected from 2010 to 2021 containing limited screened history/symptoms but demographics similar to US census as follows: Age 13-30y, Male/Female 57/43%, Asian 6%, Black 19%, Native American <1%, Pacific Islander <1%, White 68%, Other 4%;Hispanic/Non-Hispanic 27%/79%. Individual PSG site demographics reflected local populations. Prospective data collection since 2021 include >4000 uniform screening datasets (age, sex, race, ethnicity, ht, wt, screening H&P, COVID history, medications, digital ECG with results, screening outcome, and, if applicable, ECHO results). Follow up participation allowing initial cellular contact was high (avg 73%, range 51-91%/screening). Conclusion(s): Establishment of a national pediatric cardiac data warehouse enables large-scale aggregation of pediatric cardiac screening information to address deficits in the understanding and prevention of SCDY. This large real-world dataset will help establish normative data for pediatric ECGs which can facilitate development of new diagnostic tools such as machine learning and support pediatric drug and device development. [Formula presented]Copyright © 2023

Extracellular vesicles as delivery vectors to enhance killing of bacteria and viruses by airway-recruited human neutrophils

Background: Polymorphonuclear neutrophils (PMNs) recruited to the airway lumen in cystic fibrosis (CF) undergo a rapid transcriptional program, resulting in exocytosis of granules and inhibition of bacterial killing. As a result, chronic infection, feed-forward inflammation, and structural tissue damage occur. Because CF airway PMNs are also highly pinocytic, we hypothesized that we could deliver protein- and ribonucleic acid (RNA)-based therapies to modulate their function to benefit patients. We elected to use extracellular vesicles (EVs) as a delivery vector because they are highly customizable, and airway PMNs have previously been shown by our group to process and use their cargo efficiently [1]. Furthermore, our prior work on CF airway PMNs [2] led to identification of the long noncoding RNA MALAT1, the transcription factor Ehf, and the histone deacetylase/long-chain fatty deacylase HDAC11 as potential targets to modulate CF airway PMN dysfunction. Method(s): H441 human club epithelial cells were chosen for EV production because they efficiently communicate with lung-recruited primary human PMNs [1]. Relevant constructs were cloned into an expression plasmid downstream of a constitutive cytomegalovirus or U6 promoter with an additional puromycin selection cassette. EVs were generated in serumdepleted media and purified by differential centrifugation. Quality and concentration of EVs was determined by electron microscopy and nanoparticle tracking analysis and cargo content by western blot (protein) or qualitative reverse transcription polymerase chain reaction (RNA). Enhanced green fluorescent protein and messenger ribonucleic acid (mRNA) were used as controls. To test delivery to primary human PMNs, generated EVs were applied in the apical fluid of an airway transmigration model [2]. PMN activation was assessed by flow cytometry, and bacterial (PA01 and Staphylococcus aureus 8325-4) killing and viral (influenza Avirus [IAV] H1N1/PR/8/34;SARS-CoV-2/Washington) clearance assays were conducted. Result(s): To package protein, we used EV-loading motifs such as the tetraspanin CD63, Basp1 amino acids 1-9, and the palmitoylation signal of Lyn kinase. To load mRNA, a C'D box motif recognized by the RNA-binding protein L7Ae was included in the 3' untranslated region of the expressed RNA, and CD63-L7Ae was co-expressed. Airway-recruited PMNs treated with EVs containing small interfering RNAs against MALAT1 or HDAC11 showed greater ability to clear bacteria. Conversely, PMNs treated with constructs encasing MALAT1 or HDAC11 efficiently cleared IAV and SARSCoV- 2. PMNs expressing Ehf showed greater clearance of bacteria and viruses. Conclusion(s): Our findings suggest mutually exclusive roles of MALAT-1 and HDAC11 in regulating bacterial and viral clearance by airway-recruited PMNs. Expression of Ehf in airway PMNs may be a pathogen-agnostic approach to enhancing clearance by airway-recruited PMNs. Overall, our study brings proof-of-concept data for therapeutic RNA/protein transfer to airway-recruited PMNs in CF and other lung diseases and for use of EVs as a promising method for cargo delivery to these cells. It is our expectation that, by treating the immune compartment of CF airway disease, pathogentherapies, such as antibiotics will be more effective, and epithelial-targeted therapies, such as CFTR modulators, will have greater penetrance into the cell types of interest.Copyright © 2022, European Cystic Fibrosis Society. All rights reserved

Tumor agnostic comparison of immunohistochemistry and next-generation sequencing in detecting ALK fusions and assessment of ALK tyrosine kinase inhibitor efficacy

Background: Reliable methods to identify anaplastic lymphoma kinase (ALK) fusions are critical to matching patients to ALK tyrosine kinase inhibitors (TKIs) therapy, on or off trial. Various methods including FISH have been used, but immunohistochemistry (IHC) and next-generation sequencing (NGS) are most commonly employed. Evaluating the concordance of IHC and NGS is key, particularly in non-lung cancers where data is sparse. Method(s): NGS+ (MSK-IMPACT DNA hybrid capture NGS and/or RNA anchored multiplex PCR) and/or IHC+ (clone: D5F3) patients with cancers of any histology were identified as ALK+. ALK IHC was scored as negative (0), equivocal (e: 1+, 2+) or positive (3). Concordance of ALK detection (number of NGS+ and IHC+/total number of patients with NGS and IHC) was calculated. For patients with metastatic disease treated with any ALK TKI in the first-line (1L) setting, progression-free survival (PFS) was reported. Result(s): 347 ALK+ solid tumor patients were identified. As expected, the majority (96%, n=336) had lung cancer, however, 11 patients with 11 unique non-lung cancer histologies were found (3 gastrointestinal, 2 gynecologic, 1 breast, 1 thyroid, 1 primary brain tumor, 1 DLBCL, 1 PEComa, and 1 CUP). 57% had EML4-ALK fusions;36 non-EML4 ALK rearrangements were identified, including four novel fusions (PEKHA7-ALK, ZFPM2-ALK, TRIM24-ALK, ALK-MYO3B). ALK was evaluated by IHC alone in 83 patients (23.9%). The concordance rate between NGS and IHC was 85%. Among discordant cases, 11% (n=28) were IHC+/NGS-, 24% (n=63) were IHCe/NGS-, 3% (n=8) were IHCe/NGS+, and 0.4% (n=1) was IHC-/NGS+. The most frequent ALK TKIs were alectinib (n= 87, 58%) and crizotinib (n= 56, 38%). PFS on 1L ALK TKIs for patients with IHC+/NGS+ (n=134), IHC-/NGS+(n=1), IHC+/NGS- (n=8), IHCe/NGS+ (n=4), IHCe/NGS- (n=1) was 26 months, 26 months, 39 months, 41 months, 9 months respectively. Conclusion(s): In a population including multiple tumor types, NGS and IHC were highly concordant in ALK fusion detection. ALK TKI benefit may be observed in cases with discordant testing, in which only one assay detects a putative ALK fusion. Legal entity responsible for the study: The authors. Funding(s): NIH Cancer Center grant: P30CA008748. Disclosure: M.G. Kris: Financial Interests, Personal, Research Grant: Boehringer Ingelheim, National Lung Cancer Partnership, Pfizer, PUMA, Stand up to Cancer;Financial Interests, Personal, Advisory Role: Ariad, AstraZeneca, Bind Bioscience, Boehringer Ingelheim, Chug Pharma, Clovis, Covidien, Daiichi Sankyo, Esanex, Genentech;Financial Interests, Personal, Invited Speaker: Boehringer Ingelheim, Novartis, Millenium, Pfizer, Roche. A. Drilon: Financial Interests, Personal, Advisory Board: Ignyta/Genentech/Roche, Loxo/Bayer/Lilly, Takeda/Ariad/Millennium, TP Therapeutics, AstraZeneca, Pfizer, Blueprint Medicines, Helsinn, BeiGene, BerGenBio, Hengrui Therapeutics, Exelixis, Tyra Biosciences, Verastem Oncology, MORE Health, AbbVie, 14ner/Elevation Oncology, Remedica Ltd, ArcherDX, Monopteros, Novartis, EMD Serono, Melendi, Liberum, Repare RX, Amgen, Janssen, EcoR1, Monte Rosa;Financial Interests, Personal, Other, CME: Medscape, Onclive, PeerVoice, Physicians Education Resources, Targeted Oncology, Research to Practice, PeerView Institute, Paradigm Medical Communications, WebMD, MJH Life Sciences, Med Learning, Imedex, Answers in CME, Medscape, Clinical Care Options, AiCME;Financial Interests, Personal, Other, CME, Consulting: Axis;Financial Interests, Personal, Other, Consulting: Nuvalent, Merus, EPG Health, mBrace, Harborside Nexus, Ology, TouchIME, Entos, Treeline Bio, Prelude, Applied Pharmaceutical Science, Inc;Financial Interests, Personal, Invited Speaker: Chugai Pharmaceutical, Remedica Ltd, RV More;Financial Interests, Personal, Stocks/Shares: Treeline Biosciences;Financial Interests, Personal, Royalties: Wolters Kluwer;Financial Interests, Personal, Other, stocks: mBrace;Financial Interests, Institutional, Funding, Research funding: Pfizer, Exelixis, GlaxoSmithKline, Teva, Taiho, PharmaMar;Finan ial Interests, Personal, Funding, Research: Foundation Medicine;Non-Financial Interests, Personal, Member: ASCO, AACR, IASLC;Other, Personal, Other, Food/Beverage: Merck, PUMA, Merus;Other, Personal, Other, Other: Boehringer Ingelheim. All other authors have declared no conflicts of interest.Copyright © 2023 European Society for Medical Oncology

LIME Explainability on Flower Classification

Artificial Intelligence is becoming more advanced with increasing complexity in generating the predictions and as a result it is becoming more challenging for the users to understand and retrace how the algorithm is predicting the outcomes. Artificial intelligence has also been contributing in making decisions. There are many flowers in the world so the botanist scientists need help in identifying or recognizing which type of flower. The paper presents an x-ray diagnostic model and the explained with Local interpretable model-agnostic explanations LIME method. The model is trained with various COVID as well as non-COVID images. Whereas chest X-rays are segmented to extract the lungs and the model predictions are tested with perturbated images that are generated using LIME. This paper opens a wide area of research in the field of XAI. © 2022 IEEE.

And All Those Data.: Solutions by Population Management Tool Cloudcare

Background and Aims: Data usage is essential in diabetescare, but facilitating HCP's to provide patients with timely and regular (eHealth) insights is complex. We developed a brand-agnostic CE-marked, population management eHealthapplication, CloudCare, providing a 'closed-data-loop' between patients and HCP's. Method(s): It uploads insulin- and glucosedevice-data from all platforms/brands both by manual uploads (patient) or automated uploading. The system is used 8 years in Diabeter and resulted in more than 10.000 'dataloops'/year mainly from manual uploads by the patient. We analysed outcomedata and visit/contactdata. Result(s): The system helped to improve outcomes despite COVID-19, implementation of technologie and significant growth and helpt switching to remote care (Table) with 50% of children and 57% of adults reaching HbA1c below 7.5% (58mmol/mol) in 2021. To accommodate increasing data usage, automatic data-uploads and translate data to insights, we further developed the system to tracks data and offer decision-support. This allows triage-driven risk stratification of clinically relevant cases, allowing timely interventions. Data-use, frequencies of planned contacts as well as 'snoozing' periods are defined in a service level agreement with patients. Automated input from 2505 780G- and IS-CGM users, creates approx. 2200 daily datasets. Automated triaging reduces this to a workload to 20-65 relevant cases per day which are reviewed and forwarded to HCPs. Settings for triaging are flexible and temporary 'snoozing' is possible. Conclusion(s): Further evaluation studies includes clinical impact and impact on the organization including costs. We will seek to include additional clinics in the evaluation. Solutions such as CloudCare will help to integrate modern diabetes-treatment and improve outcomes.

Deep Learning Algorithms with LIME and Similarity Distance Analysis on COVID-19 Chest X-ray Dataset.

In the last few years, many types of research have been conducted on the most harmful pandemic, COVID-19. Machine learning approaches have been applied to investigate chest X-rays of COVID-19 patients in many respects. This study focuses on the deep learning algorithm from the standpoint of feature space and similarity analysis. Firstly, we utilized Local Interpretable Model-agnostic Explanations (LIME) to justify the necessity of the region of interest (ROI) process and further prepared ROI via U-Net segmentation that masked out non-lung areas of images to prevent the classifier from being distracted by irrelevant features. The experimental results were promising, with detection performance reaching an overall accuracy of 95.5%, a sensitivity of 98.4%, a precision of 94.7%, and an F1 score of 96.5% on the COVID-19 category. Secondly, we applied similarity analysis to identify outliers and further provided an objective confidence reference specific to the similarity distance to centers or boundaries of clusters while inferring. Finally, the experimental results suggested putting more effort into enhancing the low-accuracy subspace locally, which is identified by the similarity distance to the centers. The experimental results were promising, and based on those perspectives, our approach could be more flexible to deploy dedicated classifiers specific to different subspaces instead of one rigid end-to-end black box model for all feature space.

Latency-, Sync-, and Bandwidth-Agnostic Tightly-Timed Telematic and Crowdsourced Musicking Made Possible Using L2Ork Tweeter

The following paper presents L2Ork Tweeter, a new control-data-driven free and open source crowdsourced telematic musicking platform and a new interface for musical expression that deterministically addresses three of the greatest challenges associated with the telematic music medium, that of latency, sync, and bandwidth. Motivated by the COVID-19 pandemic, Tweeter's introduction in April 2020 has ensured uninterrupted operation of Virginia Tech's Linux Laptop Orchestra (L2Ork), resulting in 6 international performances over the past 18 months. In addition to enabling tightly-timed sync between clients, it also uniquely supports all stages of NIME-centric telematic musicking, from collaborative instrument design and instruction, to improvisation, composition, rehearsal, and performance, including audience participation. Tweeter is also envisioned as a prototype for the crowdsourced approach to telematic musicking. Below, the paper delves deeper into motivation, constraints, design and implementation, and the observed impact as an applied instance of a proposed paradigm-shift in telematic musicking and its newfound identity fueled by the live crowdsourced telematic music genre. © 2022, International Conference on New Interfaces for Musical Expression. All rights reserved.

A Universal 'Day Zero' Infectious Disease Testing Strategy Exploiting Crispr-Based Sample Depletion And Metagenomic Sequencing

Background. The lack of preparedness for detecting the highly infectious SARS-CoV-2 pathogen - the pathogen responsible for the COVID-19 disease - caused enormous harm to the public health, the economy and society as a whole. It took ~60 days for the first RT-PCR tests for SARS-CoV-2 infection developed by the United States Centers for Disease Control (CDC) to be made available. It then took >270 days to deploy 800,000 of these tests at a time when the estimated actual testing needs required over 6 million tests per day. Testing was therefore limited to only individuals with symptoms or individuals in close contact with confirmed positive cases. Testing strategies that can be deployed on a population scale at 'day zero' (i.e., at the time of the first reported case) are needed. Next Generation Sequencing (NGS) has such day zero capabilities with the potential to enable feasible and broad large-scale testing strategies, however it has limited detection sensitivity for low copy numbers of pathogens which may be present. Here we demonstrate that using CRISPR-Cas9 to remove abundant sequences that do not contribute to pathogen detection, NGS detection sensitivity is equivalent to RT-PCR. In addition, we show that this assay can be used for variant strain typing, co-infection detection, and individual human host response assessment - all in a single workflow using existing open-source analysis pipelines. This NGS workflow is pathogen agnostic, and therefore has the potential to radically transform how both very large-scale pandemic response and focused clinical infectious disease testing are pursued in the future. Methods. Covid positive samples with RT-PCR Ct values from 16-39 were processed through the CRISRP enhanced mNGS pipeline. Results. Sn/Sp compared to RT-PCR was 97%/100%. Strain calling concordance compared to amplicon sequencing was 100%. Co-infections from Covid positive samples were identified with high confidence. Host response signatures match the published literature. Conclusion. Applying CRISPR enhanced metagenomic NGS at Day Zero of the next pandemic can mitigate the time gap in developing approved diagnostics at population scale and potentially save lives.

Phenocomb, a Tool for Evaluating Complex Phenotypes in High-Dimension Cytometry Datasets

Background Modern cytometry can simultaneously measure dozens of markers, empowering investigation of complex phenotypes. However, manual gating relies on previous biological knowledge, and clustering/dimension-reduction tools fail to capture discrete phenotypes. Consequently, complex phenotypes with potential biological importance are often overlooked. To address this, we developed PhenoComb, an R package that allows agnostic exploration of complex phenotypes by assessing the frequencies of all marker combinations in cytometry datasets. Methods PhenoComb uses signal intensity thresholds to assign markers to discrete states (e.g. negative, low, high). As Pheno- Comb works in a memory-safe manner, time and disk space are the only constraints to the number of markers and discrete states that can be evaluated. Next, the number of cells per sample from all possible marker combinations are counted and frequencies assessed. PhenoComb provides several approaches to perform statistical comparisons, evaluate the relevance of phenotypes, and assess the independence of identified phenotypes. PhenoComb also allows users to guide analysis by adjusting several function arguments such as identifying parent populations of interest, filtering low-frequency populations, and defining a maximum marker complexity. PhenoComb is compatible with local computer or server-based use. Results In testing of PhenoComb's performance on synthetic datasets, computation on 16 markers was completed in the scale of minutes and up to 26 markers in hours. We applied PhenoComb to two publicly available datasets: an HIV flow cytometry dataset (12 markers and 421 samples) and the COVIDome CyTOF dataset (40 markers and 99 samples). In the HIV dataset, PhenoComb identified immune phenotypes associated with HIV seroconversion, including those highlighted in the original publication. In the COVID dataset, we identified several immune phenotypes with altered frequencies in infected individuals relative to healthy individuals. Conclusions PhenoComb is a unique and powerful tool for agnostically assessing phenotypes. By more fully utilizing the high-dimension data in single cell datasets, PhenoComb empowering exploratory data analysis and discovery of phenotypes for further characterization.

Phase 1/2 Study of Agent-797, an Allogeneic Invariant Natural Killer T (Inkt) Cell Therapy, in Subjects with Moderate to Severe Acute Respiratory Distress Syndrome (Ards) Secondary to Sars-Cov-2 (Covid-19)

Background AgenT-797 is a novel allogeneic iNKT cell therapy demonstrating activity in malignances and serious viral infections (i.e., SARS-CoV-2). In response to inflammatory injury, iNKTs home to critical organs, including lungs, dampen proinflammatory cytokines and protect epithelial tissues. INKTs drive response through activation of innate and adaptive immunity, recruitment/trans-activation of NK, B, and T cells, and myeloid cells via contact and soluble mediators. iNKTs represent a novel and attractive potential immunotherapy for viral ARDS. This analysis presents results from an ongoing phase 1/2 study of agenT-797 in mechanically ventilated patients with moderate to severe ARDS secondary to COVID- 19;NCT04582201. Methods As of February 2022, patients on mechanical ventilation with confirmed moderate to severe (Berlin Definition) ARDS, secondary to COVID-19 were treated with a single infusion of agenT-797 at 100, 300, or 1000 x 106 iNKT cells. Primary endpoint was safety and secondarily, time to extubation, prevention of secondary infections, persistence and alloimmunity were evaluated. Clinical benefit was defined as improvement/resolution of viral ARDS evaluated as time to extubation and survival at 30 days post-infusion. Results Twenty evaluable patients were treated with agenT-797 with a median age of 66 years (range 26-77;85% >=65y). Patients enrolled early in pandemic (pre-vaccines) and were heavily pre-treated with remdesivir, steroids and/or tocilizumab. No dose-limiting toxicities were observed. Tolerability was favorable with no cytokine release syndrome (CRS), neurotoxicity, or severe immune-related AEs. One SAE was deemed possibly related to agenT-797 (Dyspnea, Grade 4). The most frequent AEs deemed possibly related was pyrexia (grade 1;n=6). Survival was 70% (14/20) in this predominantly elderly, mechanically ventilated population. Early signals of reduction in ARDS symptoms, rapid extubation, and reduction in secondary infections were observed. AgenT-797 was detected in peripheral blood up to day 6 post-infusion, consistent with a rapid translocation from blood to tissue. Spikes in the blood during D1 and D2 showed a dose-proportional relationship, however, increased dose did not lead to prolonged peripheral persistence. Additional translational and biomarker evaluation is underway. Conclusions In patients with severe viral ARDS secondary to SARS-COV-2, agenT-797 demonstrated encouraging survival and disease mitigating benefit with a favorable tolerability profile. The deep and broad activity observed is likely attributed to iNKT cells' ability to promote viral clearance, home to the lungs, and reduce inflammation. These findings support the potential for a variant-agnostic therapy for patients with viral ARDS, a condition for which there are currently no effective therapies.

Possible applications of lung ultrasound - a Covid-19 case.

Background: Lung ultrasound is an im-portant tool for distinguishing between causes and therapies of cardiorespiratory diseases in emergency departments (ED). Aim and method: Based on a case report, the importance of point-of-care ultrasound (POCUS) in the context of emergency di-agnostics and intensive care therapy will be illustrated. Case report: A 78-year-old male presented to the ED with dyspnea und weakness. A double mRNA-Covid vaccination was completed 3 months before. His medical history revealed multiple myeloma. Using POCUS, a severe Covid-19 pneumonia could be suspected, and at the same time other differential diagnoses were ruled out. PCR confirmed a SARS-CoV-2 infection. The patient was admitted to our intensive care unit with severe Covid-19 pneumonia fol-lowed by a complicated and ultimately le-thal course. Conclusion(s): In immunocompro-mised patients, there is still a high risk of a severe and complex course despite vaccina-tion. POCUS allows evaluation of probable Covid-19 pneumonia and rapid exclusion of possible differential diagnoses. Copyright © 2022 Dustri-Verlag Dr. K. Feistle.

INTER-BRAND COVID-19 VACCINE PREFERENCES AND VACCINE HESITANCY: A SURVEY OF A CONVENIENCE SAMPLE

BACKGROUND: Delay in acceptance or refusal of vaccination despite vaccine availability comprise a continuum of attitudes known as vaccine hesitancy. To date, three COVID-19 vaccines have been granted emergency use authorization in the U.S.;yet hesitancy to accept vaccination against COVID-19 remains common. Understanding the nature of inter-brand preferences amongst7 these vaccines may help inform vaccine allocation and outreach strategies. METHODS: In April 2021, a de-identified, web-based survey was administered to a convenience sample of respondents across forty-eight states, assessing standard demographics and presence of COVID-19 vaccine brand preference. Those indicating a preference then ranked four COVID-19 vaccine brands presented in random order. Vaccine hesitancy due to brand preference was assessed as the time length for which the respondent was willing to postpone vaccination if their preferred brand of vaccine was unavailable. RESULTS: Of 1,068 respondents, 55.4% endorsed a preference for a particular COVID-19 vaccine brand. On univariate analysis, preference presence differed significantly by age (p=0.011) and religion (p=0.012). The 50-64 age group had the lowest presence of preference (47.9%) while the 18-29 (61.5%, p=0.002) age group had the highest preference presence. The religious group with the least presence of preference was Jewish (45.2%) while the Atheist/ Agnostic (60.0%, p<0.001) and Catholic (59.2%, p=0.012) groups had the highest preference presence. Upon multivariable analysis however, only age was found to be an independent predictor of preference presence (p=0.027). 45.9% (490/1,068) of all respondents would postpone vaccination if their preferred brand was unavailable, with 14.6% (156/1,068) willing to wait three weeks or longer. Willingness to postpone vaccination based on brand availability varied significantly only by religion on both univariate (p=0.022) and multivariable analysis (p=0.043), with the lowest rates of postponement among the Jewish (43.4%) and the highest among Atheists (63.0%, p<0.001) and Catholics (53.1%, p=0.073). Respondents ranked brands in one predominant order (χ2=765.64, p<0.001). Pfizer was preferred over Moderna (Z=-9.405, p<0.001), JnJ (Z=-15.545, p<0.001), and AstraZeneca (Z=-17.399, p<0.001). Moderna was preferred over JnJ (Z=-11.658, p<0.001) and AstraZeneca (Z=-16.782, p<0.001), and JnJ over AstraZeneca (Z=-10.492, p<0.001). Besides the 65+ subgroup which did not have a significant preference between Pfizer or Moderna vaccines (p=0.773), all age and religious groups had the same rank preferences with all paired comparisons similarly significant, p≤0.001. CONCLUSIONS: Age independently predicted the presence of COVID-19 vaccine brand preference while religion independently predicted vaccine hesitancy due to said preference. Further evaluation of the causes and consequences of such inter-brand preferences may inform efforts to increase vaccination among vaccine-curious individuals and facilitate progress towards herd immunity.

LISA : Enhance the explainability of medical images unifying current XAI techniques

This work proposed a unified approach to increase the explainability of the predictions made by Convolution Neural Networks (CNNs) on medical images using currently available Explainable Artificial Intelligent (XAI) techniques. This method in-cooperates multiple techniques such as LISA aka Local Interpretable Model Agnostic Explanations (LIME), integrated gradients, Anchors and Shapley Additive Explanations (SHAP) which is Shapley values-based approach to provide explanations for the predictions provided by Blackbox models. This unified method increases the confidence in the black-box model's decision to be employed in crucial applications under the supervision of human specialists. In this work, a Chest X-ray (CXR) classification model for identifying Covid-19 patients is trained using transfer learning to illustrate the applicability of XAI techniques and the unified method (LISA) to explain model predictions. To derive predictions, an image-net based Inception V2 model is utilized as the transfer learning model. © 2022 IEEE.

Oral Covid-19 Antiviral Candidate LAU-7b Demonstrates 100[percnt] Reduction of the Risk of Progression to Mechanical Ventilation and Death in Hospitalized Moderate-to-Severe Covid-19 Patients

Rationale: LAU-7b is developed as a broadly effective oral COVID-19 therapeutic targeting membrane lipids to exert dual antiviral and inflammation-controlling activity. SARS-CoV-2 reprograms host cellular lipid metabolism to favor entry and replication, a mechanism shared by all lipid-enveloped viruses. LAU-7b decreases host cell membrane lipids fluidity, inhibits de-novo cell lipogenesis, and modulates phospholipid signaling promoting resolution of inflammation. Due to its host-directed mutation-agnostic mechanism, LAU-7b utility could span across future variants, as demonstrated in-vitro against multiple SARS-CoV-2 strains and MERS-CoV. RESOLUTION, a large Phase 2/3 study evaluating LAU-7b in hospitalized COVID-19 patients, is ongoing in the US and Canada, and preliminary Phase 2 results are presented. Methods: RESOLUTION is a placebocontrolled study of oral LAU-7b, once-a-day for 14 days on top of standard of care, in hospitalized COVID-19 patients at risk of developing pulmonary complications. The Phase 2 portion of the study randomized 148 patients with moderate-to-severe COVID-19 and 84 patients in critical condition, but not on invasive ventilation. Key endpoints included proportion of patients alive and free of respiratory failure at Day 29, rates of progression to mechanical ventilation and all-causes death by Day 60, time to recovery and length of hospitalization. Results: Both study arms were highly comparable in terms of mean age, number of comorbidities and concomitant medications. LAU-7b demonstrated a 100% reduction in the risk of progressing to mechanical ventilation or death by Day 60 in moderate-to-severe COVID-19 patients. None of the 76 patients on LAU-7b required mechanical ventilation and none died, while 5 out of 72 patients on placebo progressed to mechanical ventilation (6.9% difference, p=0.025), and 4 patients died (5.6% difference, p=0.053). LAU-7b group also showed an increase of 6.9% (p=0.055) in the proportion of patients alive and free of respiratory failure at Day 29, versus placebo. Patients on LAU-7b tended to recover more rapidly and leave hospital faster. LAU-7b was well-tolerated, with safety comparable to placebo. Critically ill patients treated with LAU-7b did not show improvement over placebo, suggesting that COVID-19 patients in respiratory failure at baseline are too severely affected to benefit. Conclusion: LAU-7b showed positive results in the trial's Phase 2 portion on both survival and avoidance of mechanical ventilation in moderate-to-severe COVID-19. The confirmatory Phase 3 portion was triggered and received approval from the FDA and Health Canada, focusing on moderate-to-severe COVID-19 and using the “Proportion of patients requiring mechanical ventilation and/or death by Day 60” as primary efficacy endpoint.

Transcriptional Interaction Between RelA and cJUN

Rationale: The SARS-CoV-2 pandemic has underscored the need for novel anti-infectious strategies, including host-directed therapeutics, against existing and emerging respiratory pathogens. We have reported that an aerosolized therapeutic comprised of a Toll-like receptor (TLR)-2/6 agonist, Pam2CSK4, and a TLR-9 agonist, ODN M362, stimulate pathogen-agnostic innate immune responses in lung epithelial cells. This therapeutic (“Pam2-ODN”) promotes synergistic microbicidal activity and host survival benefit against pneumonia caused by a wide range of pathogens. Here, we study the immunomodulatory signaling mechanisms required to effect this inducible epithelial resistance. Methods: Bioinformatic analysis of transcriptional responses from human and mouse lung epithelium al cells to influenza A H1N1 or SARS-CoV-2 (GSE147507) or Pam2-ODN (GSE289984, GSE26864) were analyzed using R and IPA software to identify essential transcription factors (TFs). Lung cell population dynamics were studied for TFs related to Pam2-ODN immunomodulatory signaling using high-throughput imaging flow cytometry (IFC). Human or mouse lung epithelial cells were stimulated with PBS or Pam2-ODN and single or dual inhibitors of TFs before challeng with influenza A H3N2 (IAV) or coronavirus OC43 (CoV) to compare the epithelium-specific transcriptional control of relevant TFs using in-cell western blotting, IFC and hemagglutination for viral burdens. Results: Functional enrichment analysis revealed RelA and cJUN to be major immunomodulatory TFs of Pam2-ODN and activators of leukocyte- and epithelial-derived antiviral immune mechanisms targeting replication of influenza A and SARS-CoV-2. Cell population dynamics studied from mouse lungs confirmed activation of RelA and cJUN in CD45+, EpCAM- leukocytes and in CD45-, EpCAM+ epithelial cells, with predominant activation of the lung epithelium and none or minimal activation of structural cell populations such as fibroblasts or endothelial cells. Studies of epithelium-specific signaling in vitro revealed co-activation of RelA-(pS536) and cJun- (pS73) TFs with Pam2-ODN, and earlier onset of cJUN phosphorylation and nuclear translocation with Pam2-ODN after IAV or CoV infection. Individual or dual inhibition of RelA and/or cJUN activity in vitro disrupted the antiviral activity of Pam2-ODN of IAV infected cells. Conclusion: Pam2-ODN induces unique, pathogen-agnostic protective signaling in lung epithelial cells that involves cooperative activation of RelA and cJUN. This combined TF signaling mechanism is not observed in other structural lung cell populations after Pam2-ODN exposure. Further, the phospho-regulation dynamics of RelA and cJUN are not replicated by IAV or CoV infection alone, suggesting a novel therapeutic process that can be leveraged to protect individuals against pneumonia. (Figure Presented).

Automated, integrated and connected: Creating an interoperable laboratory medicine network in the UK

Background-Aim: In the UK and Ireland, diagnostics laboratories exchange work with other laboratories within the NHS, Public Health organisations, and private laboratories across Europe. This is because a laboratory might not have the capacity or means to test in-house, or that they need to access a specialist assay offered by an external laboratory. However, this presents an industry-wide informatics challenge as laboratories operate on a variety of information systems. Without interoperability, the exchange of work has historically relied on paper-based systems and manual processing. This, then, opens exchanged samples to the risks of human error, slow turnaround times, lack of visibility and tracking, and forfeited patient safety. The presence of this testing in every laboratory discipline, too, means that labs are, essentially, referring tests across a network of laboratories;therefore, site-to-site interfacing is not a sustainable solution, in terms of cost or maintenance, across an international network of laboratories. This presentation will discuss a globally unique solution to this challenge which exists in most healthcare systems and organisations across the world. Methods: Labgnostic, a diagnostic exchange powered by X-Lab, is a unique, global technology which enables any diagnostic laboratory to refer tests or return results to any diagnostic laboratory, anywhere in the world. It is a technological hub which is versatile, interoperable, and offers laboratories access to a network of laboratories through a single connection. By installing an interface into a laboratory’s information system, the laboratory can connect to any laboratory on the network to send or receive any test data almost instantly. Labgnostic digitises and automates the lab-to-lab referral and reporting process and removes the need for inefficient manual processes that are time-consuming, incur unnecessary costs, and result in errors. As a cost-effective, universal, and systems-agnostic solution, Labgnostic has successfully been delivering a fully interoperable laboratory medicine network at scale to the UK, Ireland and Europe since 2007. Labgnostic, which operates under the National Pathology Exchange (NPEx) in the UK, is currently used in 90% of UK NHS laboratories, which includes 100% of English and Scottish NHS laboratories, Public Health organisations, and private laboratory organisations. Results: When COVID-19 hit the UK, the UK NHS mandated that all NHS trusts were to transfer all externally referred SARS-COV-2 tests requests and results through X-Lab’s system for the faster turnaround times, safer data exchanges, and network access it offered. All UK NHS labs, through the solution, have the opportunity for capacity uplifts by transferring or receiving COVID-19 test requests to any other laboratory on the network to ensure all tests are performed in timescale required. In the UK, most laboratories work to a rapid turnaround time of under 24 hours for COVID-19 antigen tests. The project was commissioned by the NHS in mid-March and in a matter of weeks, the X-Lab team were able to deliver and implement their solution in all remaining unconnected NHS sites. Additionally, NPEx (or Labgnostic) has been the key data infrastructure for delivering COVID-19 data back to Public Health organisations, primary care facilities, and NHS bodies who deliver results to subjects and monitor data. The X-Lab team are currently working to deliver COVID-19 antibody tests at scale in the UK through their system. Conclusions: With a hub connecting diagnostic systems across the UK, Ireland and into Europe, X-Lab are now exploring use cases for this technology across other continents with laboratory business, proficiency testing providers and public sector service organisations in areas such as surveillance. This presentation will put forward the need for an integrated and interoperable laboratory medicine network throughout the world with a case study on its success in the UK and the similarities it shares with the other testing arenas.

EXPLORING THE ROLE OF CLINICAL PHARMACISTS IN AN INTEGRATED PRIMARY CARE PROGRAM

BACKGROUND: The clinical literature has shown that clinical pharmacists (CP), as part of a care team in primary practice, can produce positive clinical and cost outcomes. Furthermore, embedding a CP in a primary care setting can optimize performance in value-based payment arrangements. OBJECTIVE: The primary objective of this program description was to observe an integrated CP program, describe the role and workflow of the CP, and report preliminary program statistics, which can benefit primary care practices. METHODS: Data management used the eClinicalWorks platform. Data was collected from five different primary care locations in Houston, TX between July 2019 and June 2020, and included demographics and integrated CP activities by visit type (i.e., office, telephonic), service type (i.e., RX adherence, RX care coordination, RX consultation), and service description (i.e., duration, education, clinical chart prep, medication review, medication reconciliation). RESULTS: The integrated clinical pharmacist program incorporated a core care team made up of clinicians, clinical support staff and community care professionals. The study population included elderly adults enrolled in a Medicare Advantage (payer agnostic) plan wherein the patients were mostly Hispanic and belonged to low-income group. The prevalent conditions included type 2 diabetes mellitus, COPD, and heart disease. There was a total of 1,581 distinct encounters (office, N = 498, 31.5%;telephonic, N = 1,083, 68.5%). The average in-office appointment duration was 23.4 minutes. RX adherence (N = 309, 28.6%) was the leading activity followed by RX care coordination (N = 284, 26.2%) and RX consultation (N = 144, 13.3%). Clinical chart prep (N = 191, 38.4%) was the most common office visit activity followed by medication review (N = 145, 29.1%) and individual education (N = 109, 21.9%). The Doximity video app was used during COVID-19 pandemic to coordinate with patients and provide virtual pharmacy support, when telephonic encounters became more prevalent. CONCLUSIONS: This program description highlighted the expanded role of CP interventions integrated into the primary care workflow. In this value-based environment, primary care providers should consider CP who are well suited to offer direct patient care coordination, improve patient adherence and clinical outcomes, and implement cost-effective therapies to their practices. However, further study is needed to examine the impact of integrated pharmacy services on patient outcomes in primary care settings.

Towards Safe Medical Device Remote Control in Tele-Critical Care Applications

Introduction: The COVID-19 pandemic has advanced market awareness of the benefits of remote-controlled ventilators to reduce the exposure of healthcare workers to patients with COVID-19, enable more rapid and frequent ventilator setting adjustment, and preserve limited personal protective equipment. The US FDA permitted manufacturers to add remote monitoring and control capabilities to ventilators and infusion pumps through immediate in effect guidance [1,2]. When integrated with tele-critical care systems, remote control of medical devices allows distant clinical experts to collaborate with local clinicians to “virtually” manage the therapy of patients at hotspots. Core remote control capabilities can also be used by software applications to implement medical device control algorithms for Software as a Medical Device (SaMD). The US Army /TATRC launched the National Tele-Critical Care Network (NETCCN) to rapidly develop and deploy a platform to support COVID-19 disaster response [3]. We are investigating technical solutions, communication protocols, and safety assurance measures for integrating remote control of medical devices to the NETCCN systems. Methods: We developed an architecture and a prototype system (Figure 1) to investigate safety, security, and interoperability requirements for integration of remote control of medical devices with tele-critical care systems. The prototype system is based on OpenICE [4], an open-source interoperability platform developed by our program to transmit data and control medical devices at the patient's bedside. Customized interfaces (hardware and software) translate device proprietary protocols to ISO/IEEE 11073-10101 terminology over DDS middleware. Remote control applications of devices connected to OpenICE are implemented as either stand-alone OpenICE apps, which can be deployed inside or immediately outside the patient's room, or as web-based apps, which can be launched from any location to communicate with the OpenICE system. We refer to the former as “nearpatient remote control”, which may be at the bedside or co-located outside the room, and the latter as “far remote” control where the operator does not have physical access to the patient or medical equipment. Our prototype system uses the RTI Web Integration Service [5] to enable web-based control applications to communicate with the connected devices. Results: The generic architecture in Figure 1 is device agnostic: it can be used with critical care ventilators, IV infusion pumps, and other devices, provided that the device interfaces support remote control. As a proof of concept, we applied this architecture to a Q Core Sapphire IV infusion pump using a non-clinical control interface, and confirmed that the infusion rate could be adjusted by both near-patient and far remote (web) control applications with generally acceptable delays (3∼8 seconds from remote control action until the pump executes the change). This prototype system allows the exploration and validation of risks associated with medical device remote control in the tele-critical care context. An example of a risk identified in our study relates contention between near and far “loci of control”. Unexpected device behavior can occur if there is no mechanism to 1) explicitly prioritize loci of control that may occur simultaneously (e.g., always prioritize local control over far control to enable the local provider to regain control or prevent remote control);and 2) clearly indicate where the locus of control resides. Other risks may arise due to issues related to cybersecurity, network QoS, permission for remote control, and usability (e.g., use errors associated with far remote control due to the lack of a real-time view of the patient). We are collaborating with the AAMI InterOperability Working Group (IOWG) to share the experience and lessons learned in this effort to develop a safety standard for medical device remote control, and with other performers in the NETCCN portfolio. (Figure Presented).

Pediatric Emergency Department Visits in US Hospitals Across 2019 and 2020

Background: This research leverages Clinical Emergency Data Registry (CEDR) data collected from 2019 to 2020. Developed by ACEP, CEDR is the first Emergency Medicine (EM) specialty-wide registry to measure acute care quality, outcomes, practice patterns, and trends in emergency care. ACEP began CEDR in 2015 focusing on quantifying and enhancing quality of emergency care through collection of quality data and development of EM-specific quality metrics. CEDR has collected data for 50 million visits representing 30 million unique patients. Study Objectives: Early in the pandemic, impact on emergency departments (ED) was substantial. By May, ED visits declined nearly 40%. CDC data indicated ED visits for patients younger than 14 declined by 70%. While several studies evaluated pandemic impact on pediatric cases, most focused on children’s hospitals or particular illnesses. This research assesses the impact on pediatric visits to all US EDs over the more extended timeline of the pandemic, including total visits, visits for influenza, and visits for COVID-19. Methods: The database was queried for visit data on patients 18 years and younger from January 2019 to December 2020. Visits were counted, not individuals, so two visits by the same individual count as two visits. Percentage of visits in this cohort with a diagnosis of influenza (ICD 10 Codes: J09.X2, J10.-, J11.-) and COVID-19 (ICD 10 Code: U07.1) were calculated. Data were compared to age-agnostic ED visits and percentage of visits for COVID-19 and influenza in the general public. This data was obtained from the National Syndromic Surveillance Program (NSSP), run by the CDC. NSSP received data from 93% of US hospitals and is provided to ACEP through an agreement with the CDC. It includes data on the number of ED visits and percentage of those visits made for COVID-19 and influenza. Simple descriptive analysis was used. Results: Pediatric ED visits experienced a 72% decline starting week 11 of 2020, coinciding with the WHO characterization of COVID-19 as a pandemic. Sharp decline continues for six weeks before plateauing. However, pediatric visits remain 38% below previous visit rates (Figure 1). General population ED visits decline significantly over an identical period, however, only decrease by 42%. These ED visits make a quicker recovery, returning to 88% of previous patient volume. When reviewing influenza and COVID-19 data, general public visits indicated a considerable lack of influenza cases during COVID-19 surges. Conversely, pediatric visits do not demonstrate the same surges. Instead, data showed repeated smaller COVID-19 spikes during the year. In addition, pediatrics did not reveal influenza suppression seen in the general public. Spiking in pediatric influenza was proportional to COVID-19 spikes, offset in time (Figure 2). Conclusion: This study reveals pediatric ED visits sharply decreased in early weeks of the pandemic. Recovery of ED visits is slower in pediatric populations compared to the general population. Further, percentage of visits for influenza and COVID-19 remain low in this population and do not show typical “waves” seen in the general population. The reduction in 2020 influenza cases seen in the adult population is not reflected in the pediatric population. Influenza spiked uniquely in the pediatric population throughout 2020. Isolation of children at home during the COVID-19 pandemic may account for limited vaccination resulting in equal exposure risks. [Formula presented] [Formula presented]