The approach uses machine learning to make predictions about how patients’ conditions might progress, and whether they’d respond positively to ‘proning’ — a technique commonly used in ICUs for people with severe acute respiratory distress syndrome.
Proning is the process of turning a patient onto their front to improve oxygenation in the lungs and has been used widely throughout the pandemic, but does not help all patients. When ineffective, it can delay the start of sequential treatments such as extracorporeal membrane oxygenation (ECMO), a life-support machine that supports the heart and lungs in pumping blood around the body.
Researchers said that by using AI to analyse patients’ data daily, guidelines in clinical practice could be improved and applied to future waves of the pandemic as well as treatment of similar diseases.
“ECMO is currently the last resort for many patients, after all other less invasive interventions such as prone position have failed, but it has associated risks,” said first author Dr Brijesh Patel, clinical science lead at Imperial’s Department of Surgery & Cancer and senior intensivist at Royal Brompton Hospital.
“Over 20 per cent of all patients on a mechanical ventilator were referred to and received management advice from one of the five national ECMO centres. Patients appropriately placed early onto ECMO show better outcomes. However, only four per cent of referred patients received ECMO, which is due to a number of reasons, but one of which could have been delays in assessment of responsiveness to interventions like prone position.”
Published in the journal Intensive Care Medicine, the study was carried out by researchers from Imperial and the Royal Brompton and Harefield hospitals using retrospective data from 633 mechanically ventilated Covid-19 patients across 20 UK ICUs during the first outbreak in 2020.
Findings showed that the AI model identified factors that determined which patients’ conditions were likely to worsen and not respond to proning — these included patients with blood clots or inflammation in the lungs, lower oxygen levels, lower blood pressure and lower lactate levels.
Professor Derek Hill, professor of Medical Imaging Science at UCL and expert in medical devices, said that a 'particularly striking’ finding of Imperial’s study was that mortality was higher at the peak of the pandemic potentially due to workplace pressure resulting in poor implementation of treatment guidance.
“It is important to realise that the authors have not shown that their algorithm can be used to improve patient outcomes in subsequent waves of Covid, nor that artificial intelligence can help doctors manage heavy workloads in the peak of a pandemic,” he commented. “In practice, changes in the way patients present and are treated since the start of the Covid pandemic means any algorithms that learn from the first wave will only have limited value in helping manage later Covid surges or surges in other countries.”
The researchers are continuing to collect patient data and are currently analysing findings from the second wave. They noted that fewer drug treatments, such as steroids and tociluzimab, were available in the first wave than in the second therefore more patients may have been triaged directly to ICU for breathing support.
Professor Hill added that whilst the research provides some insights into the way that treatment decisions in the first few days in ICU impacted outcomes, a ‘great deal more validation’ would be needed before it could be used to support clinical decision making.
IEA report claims batteries are ‘changing the game’
The weight and bulk of static batteries, even domestic units, is immaterial. The IEA's trilemma is illustrated here:-...