Natural history, trajectory, and management of mechanically ventilated COVID-19 patients in the United Kingdom

• DOI: 10.1007/s00134-021-06389-z
• Publication Date: 2021-05-11T07:03:59Z
• Authors: Brijesh V. Patel; Shlomi Haar; Rhodri Handslip; Chaiyawan Auepanw...; Teresa Mei-Ling Lee; Sunil Patel; J. Alex Harston; Feargus Hosking-J...; Donna Kelly; Barnaby Sanderson; Barbara Borgatta; Kate Tatham; Ingeborg Welters; Luigi Camporota; Anthony C. Gordon; Matthieu Komorowski; David Antcliffe; John R. Prowle; Zudin Puthucheary; Aldo A. Faisal
• Keywords: United Kingdom COVID-ICU National Service Evaluation; Adult; Artificial intelligence; Original; 610; 1117 Public Health and Health Services; Prone position; 03 medical and health sciences; Mechanical ventilation; 0302 clinical medicine; Critical Care Medicine; Artificial Intelligence; General & Internal Medicine; Prone Position; EPIDEMIOLOGY; Humans; Mortality risk; Science & Technology; SARS-CoV-2; Respiration; MORTALITY; COVID-19; 1103 Clinical Sciences; Emergency & Critical Care Medicine; Respiration, Artificial; United Kingdom; 3. Good health; Artificial; PATTERNS; ARDS; Life Sciences & Biomedicine; GAS-EXCHANGE

The trajectory of mechanically ventilated patients with coronavirus disease 2019 (COVID-19) is essential for clinical decisions, yet the focus so far has been on admission characteristics without consideration dynamic course in context applied therapeutic interventions.We included adult undergoing invasive mechanical ventilation (IMV) within 48 h intensive care unit (ICU) complete data until ICU death or discharge. We examined importance factors associated progression over first week, implementation and responsiveness to interventions used acute respiratory distress syndrome (ARDS), outcome. machine learning (ML) Explainable Artificial Intelligence (XAI) methods characterise evolution parameters our visualisation tool available as a web-based widget ( https://www.CovidUK.ICU ).Data 633 adults COVID-19 who underwent IMV between 01 March 2020 31 August were analysed. Overall mortality was 43.3% highest non-resolution hypoxaemia [60.4% vs17.6%; P < 0.001; median PaO2/FiO2 day 12.3(8.9-18.4) kPa] non-response proning (69.5% vs.31.1%; 0.001). Two ML models using weeklong demonstrated an increased predictive accuracy compared (74.5% 76.3% vs 60%, respectively). XAI highlighted increasing importance, predicting mortality. Prone positioning improved oxygenation only 45% patients. A higher peak pressure (OR 1.42[1.06-1.91]; 0.05), raised component 1.71[ 1.17-2.5]; 0.01) cardiovascular 1.36 [1.04-1.75]; 0.05) sequential organ failure assessment (SOFA) score lactate 1.33 [0.99-1.79]; = 0.057) immediately prior application prone lack response. not 76% moderate hypoxemia those severe died receiving had more missed opportunities intervention [7 (3-15.5) versus 2 (0-6); 0.001]. Despite severity gas exchange deficit, most received lung-protective tidal volumes less than 8 mL/kg plateau pressures 30cmH2O. This despite systematic errors measurement height derived ideal body weight.Refractory remains major association mortality, evidence based ARDS interventions, particular positioning, implemented delayed reduced responsiveness. Real-time service evaluation techniques offer assess delivery improve protocolised evidence-based which might be improvements survival.