covLLM: Large Language Models for COVID-19 Biomedical Literature (preprint)

The COVID-19 pandemic led to 1.1 million deaths in the United States, despite the explosion of coronavirus research. These new findings are slow to translate to clinical interventions, leading to poorer patient outcomes and unnecessary deaths. One reason is that clinicians, overwhelmed by patients, struggle to keep pace with the rate of new coronavirus literature. A potential solution is developing a tool for evaluating coronavirus literature using large language models (LLMs) -- neural networks that are deployed for natural language processing. LLMs can be used to summarize and extract user-specified information. The greater availability and advancement of LLMs and pre-processed coronavirus literature databases provide the opportunity to assist clinicians in evaluating coronavirus literature through a coronavirus literature specific LLM (covLLM), a tool that directly takes an inputted research article and a user query to return an answer. Using the COVID-19 Open Research Dataset (CORD-19), we produced two datasets: (1) synCovid, which uses a combination of handwritten prompts and synthetic prompts generated using OpenAI, and (2) real abstracts, which contains abstract and title pairs. covLLM was trained with LLaMA 7B as a baseline model to produce three models trained on (1) the Alpaca and synCovid datasets, (2) the synCovid dataset, and (3) the synCovid and real abstract datasets. These models were evaluated by two human evaluators and ChatGPT. Results demonstrate that training covLLM on the synCovid and abstract pairs datasets performs competitively with ChatGPT and outperforms covLLM trained primarily using the Alpaca dataset.

EBUS audit: assessing the diagnostic performance of the service at UHL

Introduction: Endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) is an important diagnostic procedure in the lung cancer pathway. False-negative or inadequate sampling can lead to inaccurate staging or delay in diagnosis. This study was conducted to assess the performance of the Cancer EBUS service at a tertiary hospital. Method(s): We conducted a retrospective analysis of patients who underwent EBUS-TBNA for suspected cancer between 1st June 2021 to 31st May 2022. Request forms, CT reports, EBUS reports and pathology reports were reviewed for analysis. Result(s): 205 EBUS-TBNA procedures were performed. All patients had an appropriate staging CT prior to procedure. The mean time to test was 10.5 days (7.4). 77 (38%) had tests within 7 days of request. 293 lymph nodes and 10 mass lesions were sampled. The mean time to pathological results being received was 2.9 days (1.8). Final histology showed 39 (19%) cases of lung adenocarcinomas, 3 (1%) lung non-small cell carcinomas, 25 (12%) lung squamous cell carcinomas, 25 (12%) small cell cancers, 4 (2%) lung NOS, 3 (1%) pulmonary carcinoid, 2 (1%) lymphoma, 12 (6%) other cancers, 12 granulomata and 1 thyroid tissue (6%). 43 (21%) cases showed lymphoid tissue and 28 (14%) were reported as inadequate. No samples were taken in 8 cases (4%). Adequate tissue for predictive marker testing was available in 93% (66) of cases of non-small cell lung cancer (NSCLC). Complications were encountered in 9 cases (4%). Only 3 cases (1.5%) required any form of intervention. [Figure presented] Conclusion(s): Our data provides assurance of safety while also highlighting specific areas for attention regarding performance and time to test that can be addressed and our sensitivity was comparable to national standards. The increased waiting times may be partly related to COVID-19 precautions and will require reauditing at a later date. Disclosure: No significant relationships.Copyright © 2023 Elsevier B.V.

Palliative care for Continuous Positive Airway Pressure (CPAP) withdrawal in patients with COVID-19

Introduction: Patients with COVID-19 who require CPAP as ceiling of care have a high mortality. Evidence is lacking for appropriate end of life care for these patients and withdrawal from high CPAP pressures can be distressing for patients and their families. Method(s): We retrospectively reviewed case notes for 36 patients with COVID-19 pneumonitis who were not suitable for intubation and required withdrawal of CPAP at Colchester Hospital, UK from 29th April, 2020 to 13th February, 2021. Result(s): Mean age was 74.14 (SD 9.48) years, 14 patients were female. Decision to withdraw was discussed with family in 91.7% of cases and with the patient in 50% of cases. A family member was present during withdrawal in 38.9% of cases. All patients after CPAP withdrawal died. Mean time to death since withdrawal (known in 30 patients) was 26.06 (SD 43.84) hours. Reasons for withdrawal included patient deterioration (55.6%), patient discomfort (36.1%) and patient wish (8.3%). Sedative drugs were often needed to achieve comfort, high doses were required in some cases. Comfort was achieved with midazolam, morphine and levomepromazine in the majority, however in one case, phenobarbitone was also required. Total doses used from time of decision to withdraw CPAP to death, are summarised in table 1. Conclusion(s): Optimum palliative care is vital for looking after these patients. In our experience, some patients required higher than usual doses of sedation whilst CPAP was withdrawn. (Table Presented).

Maskbot: Indoor Traffic Control

The transmission of Covid-19 continues to be a serious concern of many institutions in industrialized countries. However, circumstances dictate that on-site work must continue in education and industry. How authorities address the pandemic is a subject of controversy, and enforcement of Covid-19 protocols locally by everyday workers and educators can be uncomfortable, if not dangerous. The Maskbot project aims to automate two of these protocols: people should wear facemasks in public areas to minimize exhalation of water droplets in the air, and people should quarantine if they are running a body temperature above nominal values. With Maskbot, enforcement of these protocols take the shape of indoor traffic control of incoming and outgoing movement. Incoming traffic is audited with a Raspberry Pi B+ loaded with a model that allows it to analyze faces for masks with the feed coming from a USB camera. Additionally, incoming traffic is checked for high body temperature with an infrared thermometer. If an unmasked or irregularly high temperature status is received, the entrance gate will block incoming traffic with barrier arms attached to two parallax servos, controlled by a linked Arduino Uno. Furthermore, these undesired states will result in an alarm going off, an LCD informing the passerby that they are not to proceed, and a picture being taken of the passerby to be sent to a building manager email. Desired states will result in a prompt to proceed and raise barrier arms. Outgoing traffic is managed by an exit gate, which raises its barrier arms on command when a leaving pedestrian waves their hand in front of an ultrasonic sensor. © American Society for Engineering Education, 2022.

COVID-19 mortality rates in a district general respiratory support unit

P214 Table 1Mortality by groups Age groups 30–39 40–49 50–59 60–69 70–79 >80 Mortality (%) 0 40 50 71.4 79.5 100 CFS 1 2 3 4 5 6 Mortality (%) 20 43 79 77.3 87.5 100 Number of comorbidities 0 1 2 3 4 Mortality (%) 42.9 46.1 76.9 80 84.6 Time to admission to RSU (days) Less than 1 1-2 3-7 8 or more Mortality (%) 63.2 64.7 81.5 81.8 ConclusionThese data demonstrate the high mortality in patients with respiratory failure secondary to COVID-19 pneumonitis admitted to our RSU, particularly in those who were not for escalation to ICU. Increasing age, number of comorbidities and CFS were associated with treatment failure as well as time between presentation and admission to RSU. Careful patient selection with consideration of these factors is vital when identifying patients appropriate for respiratory support.

Determining the importance of high-reliability healthcare organizations during the Covid-19 pandemic: Evidence from healthcare workers in Turkey

The Covid-19 pandemic has caused changes in the social and economic environments for healthcare. Particularly, to avoid spreading the Coronavirus pandemic, release the stress among healthcare workers, and make them work effectively during the epidemic, high-reliability healthcare organizations give great importance to the improvement of their functions. This study aims to show the importance of high-reliability healthcare organizations comparing their effectiveness during a pandemic by applied qualitative research method with many statistical analyses. In order to achieve the aim of the study, a Likert scale survey technique is used to collect the data by using an online survey. 280 healthcare workers filled the survey from January 17, 2021, to February 22, 2021. Based on the outcomes of the analyses, it has been found that such functions as shared knowledge pattern, provision of self-care, awareness of the coronavirus consequences at the workplace of high-reliability healthcare organizations have a positive and significant relationship at p < 0.01 level with taken appropriate measures against coronavirus variable. Self-awareness of organizational role, organizational resources to provide safety, flexibility of work, environmental safety, and collective mindfulness do not have any relationship with the appropriate measures against Covid-19 variable. This outcome indicates that shared knowledge pattern, provision of self-care, and awareness of the coronavirus consequences at the workplace have a more important role in combating Covid-19 in high-reliability healthcare organizations. © Imran Sarihasan, Judit Oláh, Main Al-Dalahmeh, Allam Yousuf, Krisztina Dajnoki, 2021