Scientists say they have identified six different “types” of Covid-19, each based on a “particular cluster of symptoms”.
Their findings are based on data gathered from King’s College London’s COVID Symptom Study app.
The researchers used the data to develop a model to predict which patients are most likely to require hospitalisation and breathing support, based on their earliest symptoms. The team said these findings could have “important implications for care and monitoring of people who are most vulnerable to severe Covid-19”, ahead of a second wave of coronavirus infections.
According to the researchers, most people who require breathing support come to hospital around 13 days after their first symptoms, so knowing those who are at higher risk of severe infection, as early as day five, could help save lives.
Study author Dr Claire Steves, from King’s College London, added:
“If you can predict who these people are at day five, you have time to give them support and early interventions such as monitoring blood oxygen and sugar levels, and ensuring they are properly hydrated – simple care that could be given at home, preventing hospitalisations and saving lives.”
Continuous cough, fever and loss of smell are recognised as the three main symptoms of the disease caused by the Sars-Cov-2 virus.
If you can predict who these people are at day five, you have time to give them support and early interventions such as monitoring blood oxygen and sugar levels, and ensuring they are properly hydrated
But the scientists said data from their app showed a wide range of symptoms associated with Covid-19 such as headaches, muscle pains, fatigue, diarrhoea, confusion, loss of appetite and shortness of breath.
The team used machine learning, a type of artificial intelligence, to see whether particular symptoms tend to appear together and whether that had an effect on the progression of the disease.
Analysis was made based on the data gathered from around 1,600 users in the UK and US, between March and April, who had a confirmed Covid-19 diagnosis. The results showed six specific clusters of symptoms, which according to the researchers, represented six distinct types of Covid-19.
These clusters include:
– Flu-like with no fever: Headache, loss of smell, muscle pains, cough, sore throat, chest pain, no fever
– Flu-like with fever: Headache, loss of smell, cough, sore throat, hoarseness, fever, loss of appetite
– Gastrointestinal: Headache, loss of smell, loss of appetite, diarrhoea, sore throat, chest pain, no cough
– Severe level 1, fatigue: Headache, loss of smell, cough, fever, hoarseness, chest pain, fatigue
– Severe level 2, confusion: Headache, loss of smell, loss of appetite, cough, fever, hoarseness, sore throat, chest pain, fatigue, confusion, muscle pain
– Severe level 3, abdominal and respiratory: Headache, loss of smell, loss of appetite, cough, fever, hoarseness, sore throat, chest pain, fatigue, confusion, muscle pain, shortness of breath, diarrhoea, abdominal pain.
The researchers said they recently identified skin rash as another key symptom of Covid-19 but it was not recognised as a symptom during the time when the data was gathered for their analysis – so it is currently unknown how skin rashes map on to these six clusters.
Being able to gather big datasets through the app and apply machine learning to them is having a profound impact on our understanding of the extent and impact of Covid-19, and human health more widely
The researchers said they tested their machine learning technology on another dataset, which included 1,000 users in the UK, US and Sweden, who had logged their symptoms during May, and found similar results. The team also found that those who belonged to severe level 1, 2 and 3 cluster types were more likely to be older and frailer, be overweight, and have underlying conditions such as diabetes or lung disease.
The researchers then developed a model to predict which cluster a patient falls into, and their risk of requiring hospitalisation and breathing support, based purely on age, sex, BMI and pre-existing conditions alone. They believe prediction tools like these could provide an early warning sign as to who is most likely to need more intensive care, based on their early symptoms.
Sebastien Ourselin, professor of healthcare engineering at King’s College London and senior author on the study, said:
“Being able to gather big datasets through the app and apply machine learning to them is having a profound impact on our understanding of the extent and impact of Covid-19, and human health more widely.”
The study is published in a pre-print server called medRxiv, and is yet to be peer reviewed.