The new study provides predictive analysis of six different symptoms that patients with COVID-19 may experience.

To predict the development of more severe respiratory diseases later in life, the team at King’s College London analysed existing diagnostic records. Using machine learning techniques, the researchers concluded six different symptoms based on data from different ages, genders, medical conditions, and five days before and after onset. Among those who need breathing support, they believe the accuracy of the prediction is close to 80%.

The new study provides predictive analysis of six different symptoms that patients with COVID-19 may experience.

(From King’s College London)

The study has now been published on medRxiv’s preprinted website, meaning it has not yet been peer-reviewed. Even so, Professor Tim Spector told the Guardian: “Early interventions can help increase patients’ chances of survival while avoiding unnecessary clinical resource appropriation.”

The team is known to have screened more than 4 million mobile app data, involving 1,653 users who tested positive for COVID-19. It reported its symptoms and provided an update on its health status. It is known that 383 people went to the hospital at least once, and 107 of them needed oxygen or ventilator support.

After putting on a machine learning algorithm, the team looked at the symptoms in 14 and finally determined six groups of different COVID-19 patients to infer what could provide a higher predictable accuracy.

Group 1 (462 cases): With symptoms of the upper respiratory tract such as persistent cough and muscle pain, about 1.5% of patients need edire support, and 16% of them have been to the hospital at least once.

Second group (315 cases): high frequency of appetiteless and fever symptoms, 4.4% of patients need breathing support, 17.5% of patients have been to the hospital at least once.

Group 3 (216 cases): Gastrointestinal symptoms such as diarrhea, but fewer other symptoms. 3.7% of patients in this group need breathing support and 24% have been to the hospital at least once.

Group 4 (280 cases): Early symptoms of severe fatigue, persistent chest pain and cough, 8.6% of patients needed respiratory support, and 23.6% of patients visited the hospital at least once.

Group 5 (213 cases): Symptoms such as mental disorders, loss of appetite, severe fatigue, 9.9% of patients needed respiratory treatment and 24.6% of patients visited the hospital at least once.

Group 6 (167 cases): Significant respiratory distress, shortness of heart, chest pain, blurred consciousness, fatigue, gastrointestinal symptoms. Up to 20 percent of patients in this group need breathing support, and 45.5 percent have been to the hospital at least once.

The team noted that the first two groups of COVID-19 patients had less symptoms. In addition, they repeated the work using user data from other 1047 apps, and added symptom descriptions such as headache and reduced taste sensitivity, with varying durations of symptoms for different patients.

But by compiling data on symptom reports for the first five days, as well as information about the patient’s age, gender, and other medical conditions, the team claims to be able to achieve a fairly high rate of prediction accuracy, such as 79% of cases requiring respiratory support. However, if you focus only on the patient’s individual symptoms, the prediction accuracy is or less than 70%.