COVID-19 Intubation to fight Respiratory Problems

Doctor examining patient's lung

We hear a lot about intubation lately when discussing patients with COVID-19-related respiratory problems. Yet the procedure isn’t particular to COVID. Some people are too vulnerable to breathe autonomously. For instance, they have chronic obstructive pulmonary disease (COPD) or pneumonia. 

How does intubation work?

The medical staff intubates a patient by placing an endotracheal tube into a person’s trachea through the mouth or nose. It helps open up the patient’s airways to secure breathing. It’s simple and straightforward airway management to prevent obstruction. Intubation occurs often, for instance, when someone is undergoing anesthesia in the operating room or in emergencies when someone is too ill or injured to breathe autonomously. 

The tracheal tube is placed into the windpipe. It terminates just before the place where the trachea heads into the lungs. The flexible plastic tube is then combined with a mechanical ventilator to fight respiratory distress. Clinicians also intubate patients with a breathing tube to keep them ventilated in critical care.

Intubations become necessary when you reach a point where your lungs cannot provide your body with oxygen. It also occurs when your body can’t get rid of the carbon dioxide, the oxygen saturation becomes elevated. 

Clinicians call this respiratory problem acute respiratory distress. At this stage, rapid sequence intubation becomes the only choice to save your life. The ventilators and tracheal tube typically help secure the airway and enable medical staff “to give air faster.”

COVID-19 patients with respiratory problems are not all undergoing tracheal intubation and figures vary depending on dates and places. 12% of patients experienced mechanical ventilation through intubation in March 2020. 

The reference here is the JAMA case series concerning 5700 COVID-19 patients in New York. A similar study was conducted simultaneously in Atlanta, Georgia, on 217 COVID-19 patients. The medical staff practiced this emergency medical procedure to ensure patient airway in 35% of cases. 

Doctors use intubation in emergency departments when facing a life or death condition, and all else has failed. They secure the airway and thus oxygenate the patients. In a way, the doctors force the respirations. The clinicians often administer antiviral drug Remdesivir, and anticoagulants to inhibit blood clots in COVID patients. 

They may also receive supplemental oxygen and convalescent plasma, a therapy that contains SARS-CoV-2 antibodies. Sometimes a patient’s state worsens to cause complications. When they result in an inability to breathe regularly, doctors and paramedics will move ahead. They will enforce advanced airway management.

Doctors can intubate some people and provide ventilatory assistance for a couple of days. Others can be intubated and mechanically supported for weeks. Many people believe that intubation is related to morbidity, but that’s not necessarily the case. 

Statistically, most people who are intubated for COVID-19 will eventually get extubated and survive. ICU does not mean death. And intubation means airway protection based on sound medical advice, not death. 

The JAMA case series showed that the death rate among people who bore mechanical ventilation through intubation was 24,5%. While the figure appears to be important, it also means that approximately 3/4 of people survived.

Respiratory Problems Complications Analyzed with TADA

Is it possible to dig deeper? Can we analyze who, among COVID-19 patients and non COVID, experience respiratory distress syndrome, and get intubated?

As a data scientist team, we have studied two tracks to understand who might be more prone to experience respiratory disease as in our two first analyses about contamination and ICU. The latter consists of using a Machine Learning tool fine-tuned to perform well on small amounts of data. 

The database we have used is made publicly available by the Mexican government. It is used in numerous research studies and papers. One of these studies is an article by the medical researcher Omar Yaxmehen Bello-Chavolla and his team. The records in this database are inputs from Mexican hospitals. 

We have analyzed the overall database focusing on patients with respiratory conditions leading to endotracheal intubation (with and without COVID). 

Statistically, among the overall hospital population considered, 13% suffered from airway obstruction and were intubated, 87% did not. The following table provides the repartition of men and women with regards to intubation. It shows that men are intubated twice as often as women. 

MenWomen
Not Intubated49%38%
Intubated8.3%4.9%

The next significant criterion, according to statistics, is pneumonia. Among the non-intubated patients, 38% have pneumonia, and 48% don’t. That’s a 1:1.2 ratio. However, 3% of intubated people do not have pneumonia when 10% do; that’s a 1:3 ratio. 

Among the other criteria reported in this database are: diabetes, asthma, pregnancy, obesity, cardiovascular disease. The percentage of people who experience breathing problems leading to intubation is the same with and without the above conditions. In a nutshell, according to statistics, the two key elements which induce respiratory problems are gender and pneumonia.

We have run our Machine Learning tool, TADA, on a subset of 10 000 samples representative from this database. The accuracy measured is 60% with a sensitivity of 65%.

The top three criteria for being intubated according to TADA are: being over 65, having pneumonia, and being a contact case. So gender does not count so much for TADA. But the most exciting part is the model’s false-negative rate, which was small. By using the above criteria for prediction, this rate was: 131 out of 3 000. 

It means that 131 times out of 10 000, TADA predicted “not intubated,” and the person was indeed intubated. This low number is a good figure in a medical environment. It is better to overpredict difficulty breathing than to underpredict it. And, for instance, to encounter a shortage of ventilators as a consequence.  

We have also experienced running TADA on a reduced database. It contained only COVID patients, and the pneumonia information was discarded. The accuracy decreased a little to 55%, and so did the sensitivity of the resulting model with 54%.

However, the criteria that TADA elected to build the model were obesity, age, and asthma. It is a fascinating result. The key factors impacting respiratory conditions in patients with COVID are entirely different from what the statistics seem to indicate at first sight. 

We have set aside pneumonia. It was present in 90% of patients with COVID and experiencing respiratory problems, i.e., intubated. Therefore it was overshadowing other pertinent factors. In this case, too, the number of false negatives is small: 209.

Respiratory Problems, Covid and Intubation in Short

People sick with COVID are more likely to suffer respiratory illnesses. They endure more often respiratory problems and have to be intubated to ensure their upper respiratory tracts are unobstructed. While in the general population, age, pneumonia, and being a contact case were significant risk factors. 

In the COVID population, TADA’s top risk factors are obesity, age, and asthma. Quite a different story. That’s the whole point of using an interpretable tool. It provides the medical researchers with insights out of a list of figures. 

References

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7255393/

https://www.prevention.com/health/a33297904/what-is-intubation/

https://jamanetwork.com/journals/jama/fullarticle/2765184

https://www.webmd.com/lung/intubation-explained

Share
Share on linkedin
Share on twitter
Share on facebook

Start making sense of  your data

Test easily TADA with our test data here

You might also like...

Doctor studing on pulmonary disease

MyDataModels Wins the EUREKA International Call for Projects

MyDataModels & Vertical M2M Bring AIoT One Step Further

Semi-supervised learning for document classification

Sorting Algorithms

Multiclass Classification: Sorting Algorithms