The effect of Respiratory Support in COVID-19; a causal inference study (the RS-COVID Study) 


UHB, UoB, The Turing, UCL, UCLH, Oxford, LSHTM 

Lay Summary: 

In March 2020 the World Health Organisation (WHO) declared the disease COVID-19, caused by the virus SARS-CoV-2, a pandemic. Although most people infected with the virus have no, or very mild symptoms, in 20% of those infected, the symptoms are bad enough to require medical attention and admission to hospital, with the most severely affected requiring care on the Intensive Care Unit (ICU). 


When patients are admitted to hospital with COVID-19, many have difficulty breathing and 

require a machine to support their breathing. There are two main ways to do this. A tight-fitting face mask (called CPAP or “Continuous Positive Airway Pressure”) can be placed over the patient’s mouth and nose, and pressure is applied to help oxygen into the lungs.  


Alternatively, the patient is placed in a medical coma (made very sleepy with a general anaesthetic) and a breathing tube is placed through the mouth and into the windpipe. This is called IMV or Invasive Mechanical Ventilation.  Here, the patient is kept asleep on an intensive care unit and the breathing tube is connected to a ventilator which “breathes” for the patient.  


CPAP can be performed with fewer resources, on a normal hospital ward and with patients who are awake so can continue to engage with their treatments, nursing care and speak to family members. CPAP is associated with less side effects than IMV. However, patients may need to remain on CPAP for considerable time, which can be exhausting. A theoretical risk is that patients could deteriorate on CPAP and do worse than if they had received intubation and IMV from the outset.  


In COVID-19, both approaches may require many days or weeks of treatment. Each type of breathing support comes with its own risks and benefits. The face mask approach requires continued effort from the patient, which could exhaust the patient and do more harm 

in the long run. But it has the advantage that the patient can remain awake, and so continue 

to engage with their treatment, nursing care, and remain in contact with family. Often the face mask is used to try and prevent the need for a breathing tube.  The breathing tube means the patient is now asleep in a medical coma, and so is unable to communicate 

easily with family or healthcare staff. However, this more invasive treatment allows the patient to rest while the lungs heal. 


As COVID-19 is a new disease, hospitals across the world had no guidelines to say which treatment was better in severe COVID-19, the mask (CPAP) or the breathing tube or to direct what kind of patient may benefit from one approach or the other. 


The RS-COVID study has been designed to help understand the role of breathing support in COVID-19, to ensure patients are offered the best treatment for them. 


This study will seek to answer the following questions: 

• Is one particular method better than the other? 

• Does using the oxygen face mask lead to worse outcomes as a result of delaying the use of a breathing tube? 

• Does delaying using a breathing tube ultimately lead to more harm overall? 

• Which types of people benefit more from a particular type of therapy? For example, do older patients benefit more from the facemask approach? 


This study uses data that is already collected as part of routine care, to try and answer the 

question of which approach is better, for whom, and if delays in the use of a breathing tube are causing harm. 


Within DECOVID the two primary collaborating hospitals used different approaches for providing respiratory support. One used the breathing tube and very little CPAP, the other used both.  This allows us to conduct a study to understand which therapy (CPAP or IMV) is most beneficial, for whom, and over what timescale.  


This study will make use of statistical techniques specifically designed to answer these sorts of complex questions.   


The results will be shared in open access journals, so that as many people as possible can benefit from what we learn.  All analytical code (the computer code we use to analyse the data) will also be made available any researchers, to ensure that our research follows “FAIR” principles (Findable, Accessible, Interoperable and Reusable). 

Scientific Summary: 

In its most severe form, COVID-19 leads to hospitalisation of patients with an acute severe respiratory distress (ARDS) like syndrome. Despite newly approved specific therapies in the most severe cases, the mainstay of treatment has been supportive in nature; oxygen supplementation and ventilatory support.  


For patients who deteriorate beyond the capability of simple application of oxygen, two broad ventilator strategies have emerged. The first is early non-invasive support with continuous positive airway pressure (CPAP) delivered via a tight-fitting face mask. The second is early invasive mechanical ventilation (IMV) via the introduction of an endotracheal tube into the lungs under general anaesthetic. There are theoretical risks and benefits to both strategies but very little evidence to guide treatment decision making.  


Many hospitals in the UK and globally have implemented competing strategies. 

CPAP can be performed with fewer resources, on a normal ward and with ambulant patients. Patients remain conscious, and so can continue to engage with their treatment, nursing care and family communication. These patients are not exposed to the myriad potential harms associated with intensive care, including medically induced coma (sedation), immobility, or the potential infection from invasive lines. However, patients may need to remain on CPAP for considerable time, which can be exhausting. A theoretical risk is that patients could deteriorate on CPAP and do worse than if they had received intubation and IMV from the outset. Endotracheal intubation and IMV is more resource intensive, and must be delivered in an intensive care unit. IMV is often accompanied by many interventions that may themselves cause harm, as outlined above. Intubation itself is not risk free, as instrumentation of the airway can and does result in fatality. However, IMV can support patients with a much greater degree of disease severity, and for longer than CPAP. Patients on IMV can in theory rest while the lungs heal. 


The primary hospitals of DECOVID undertook different strategies. Both hospitals provided protocolised care, with one favouring a CPAP first approach, and the other preceding directly to endotracheal intubation. This analysis will use advanced causal inference methodologies to identify if there is any benefit or harm caused by either respiratory strategy. 



• To investigate the causal effects of Continuous Positive Airway Pressure (CPAP) versus 

invasive mechanical ventilation (IMV) on COVID-19 outcomes (primary: mortality). 

• To determine which characteristics are associated with treatment initiation, escalation or 


• To determine whether there are disease phenotypes and patient characteristics that affect the degree of benefit from either therapy. 


Research Questions: 

1.       Descriptive analysis of the numbers of patients in receipt of each treatment and their outcomes, the trajectories they follow, and the characteristics at the start and change of respiratory therapy. 

2.       Which patient characteristics are associated with the need for oxygen-supplementation and CPAP/IMV initiation. Which are associated with the need to progress treatment? 

3.       What are the causal effects on patient outcomes of different treatment regimes? Static regimes, where the treatment pathway is fixed on external criteria, include “always treat”: the patient is placed on IMV or CPAP at admission, “treat in the first 12 hours of hospital admission”, or “never treat”: the patient is maintained solely on simple oxygen, and receives no higher respiratory therapy. Dynamic regimes, where the treatment pathway is determined by pre-defined ‘rules’ such that treatment initiation or escalation is done on the basis of a patient’s own physiological markers (absolute level and/or deterioration). 

4.       Can time-updated patient information be used to provide time-updated predictions of 

           patients’ outcome (known as “Dynamic predictions”)? Time updated patient information used will include history of vital signs, interventions and medical background, and signature type technology. 

5.       Do the effects of CPAP or IMV strategies differ across patient subgroups characterised in terms of age, comorbidities, COVID severity, inflammatory phenotype or compliance with therapy.