-->

Encrypting your link and protect the link from viruses, malware, thief, etc! Made your link safe to visit.

Machine learning model from the largest US COVID-19 dataset predicts disease severity

The results of a central repository of COVID-19 records, which was built last year, are beginning to show some signs. A new paper is the first to be published. This repository contains the largest collection of COVID-19 records and was created by data specialists and researchers last year to make sense of COVID-19.

The study was published in JAMA Network Open and examined risk factors for severe COVID-19 cases. It also tracked the progression of the disease over the years. Based on data collected in the hospital's first day, machine learning models were developed to predict which patients will develop severe disease.

The National COVID Cohort Collaborative Data Enclave or N3C was used by the research team to allow it to access hundreds of thousands of patient records. The study included data from 34 medical centres and more than 1 million adults. It included 174,568 people who were positive for COVID-19, and 1,133.848 who were negative. The records range from January 2020 to December 2020.

This analysis shows that COVID-19 treatment has changed as doctors have tried new methods and gained more knowledge. After being promoted by Donald Trump, the anti-malaria drug, hydroxychloroquine was found to be ineffective. The proportion of patients who received it dropped to almost zero by May 2020. After studies showing that it could increase survival rates, the use of the steroid Dexamethasone jumped in June.

It was also found that patients with COVID-19 have had their survival rates increase over the past year. 16 percent of COVID-19 patients admitted to the hospital died between March and April. This dropped to less than 9 percent in September and October.

Patients with higher heart rates, breathing rates and temperatures at the time they arrived at the hospital were more likely need to have ventilation. They also had a higher chance of dying. More severe cases were also associated with abnormal white blood cells, inflammation, blood acidity and kidney function. The machine learning models were built by the research team using these data points and other data to predict which patients will become seriously ill. The authors suggested that the models could be used with additional testing to help in the development of decision-making tools.

Since the beginning of the pandemic, researchers have been studying the COVID-19 trajectory. This study draws on a wide range of data sources, so it is not limited to one state or hospital. Researchers in the US are limited to the study of medical records from patients within the hospitals where they work. This means that they may not be able access all the records necessary to complete their studies and can't easily verify if they would work elsewhere.

N3C, which draws together records from many institutions, bypasses these limitations. N3C currently has data from 73 institutions and records from more than 2 million COVID-19 people. There are more than 200 research projects that use the data. These include studies on risk factors and pregnancy effects of COVID-19 relapse. It is not perfect. Standardizing information across hospitals can be difficult and may not provide complete data for many patients.

However, such a vast amount of data can be invaluable. Researchers can use the data to conduct studies they might not be able to do with their institution's resources. Elaine Hill, a University of Rochester health economist who specializes in pregnancy research, said that researchers are now using this resource. She said, "It allows us to shed light on matters we wouldn't otherwise be able."

ST