To Ventilate or Not To Ventilate: a tool to help prioritize patients

Simon Duchesne's team is developing a decision-support algorithm to support the clinicians who might be called upon to make this important decision.

By: Jean Hamann, translated by the CERVO team

Read the original article by Jean Hamann on the website: ULaval Nouvelles (in French)

This news story has been re-published in Le Soleil newspaper (in French)

Should the worst-case scenario occur and the number of people with COVID-19 requiring mechanical ventilation exceeds the number of ventilators available, committees will determine which patients should be connected as a priority. Simon Duchesne has been thinking of the people that will have to make these difficult decisions these past few days.

"No one here has ever had to face such an eventuality. We learn in a crisis context and it's a lot to ask in the heat of the moment. We lack data on which to base decisions like these. Let us be clear, the tool we are developing will not decide who lives or dies. Its purpose is to provide new information that could help committees make these decisions", says the Faculty of Medicine professor and researcher at the CERVO Research Centre of the CIUSSS de la Capitale-Nationale.

A biomedical engineer by training, Professor Duchesne has acquired expertise in the art of making medical images, particularly brain scans, speak for themselves. The pandemic has led him to transpose this expertise to the problems that now arise in intensive care units. "We posited that chest X-rays contain information that can tell us how COVID-19 will evolve in a patient," he says.

The tool he and his collaborators are working on has two goals. The first is to predict at the time of admission to the emergency department or intensive care unit whether a patient will require mechanical ventilation within 24 hours. "We now know that the condition of a patient with COVID-19 can degenerate very quickly. The earlier ventilation is started, the better the chances of survival. Therefore, we may be able to act more quickly."

The second objective is to determine when mechanical ventilation has the best chance of success. "If there is a shortage of ventilators, the plan is to give priority to those patients who are most likely to survive. It may seem simple, but it's not," says Duchesne. There is little evidence on the prognosis for the success of a mechanical ventilation intervention and this data has not been calibrated for COVID-19. Their tool would fill this gap.

Image source ULaval Nouvelles

Simon Duchesne and collaborators believe chest X-rays can help predict how COVID-19 will evolve in patients

225,000 chest x-rays to the rescue

To create this algorithm, Duchesne and his collaborators first used 225,000 x-rays of people suffering from pneumonia caused by a bacterial or viral infection. "The algorithm "learned" to associate particularities of these lung images with the description given by radiology experts," he says. The algorithm was then tested on about 100 chest X-rays, from the public domain, showing patients with COVID-19. "Our system now knows what features of a chest x-ray it needs to monitor for pneumonia in a person with COVID-19. The features that will determine the degree of success of ventilation are probably so subtle that they would escape the naked eye. We hope that the algorithm will be able to detect them," explains Professor Duchesne.

The next step, which will be carried out in conjunction with seven major hospital centres in Quebec, is to combine information from chest X-rays with clinical data, vital signs, laboratory test results, interventions and their outcomes in the first hundreds of Quebec patients who had COVID-19. "Some of these patients were placed on mechanical ventilation and others were not. Some survived, while others unfortunately died. All of these situations will allow us to establish our prediction algorithm. The more cases we have, the better we can improve it."

Simon Duchesne is communicating with public health and safety officials who will write the guide that committees will use in the event of a shortage of artificial respirators. He anticipates that his tool will be ready within one or two weeks, when the number of hospitalizations in Quebec will reach a peak. "Ideally, there will be enough artificial respirators for all the patients who need them and we will never have to make such cruel decisions," he hopes.