According to a new study, the number of deaths from COVID-19 among hospitalized patients has nearly halved after rising in the early months of the pandemic.
Researchers from the University of Colorado and Johns Hopkins University analyzed data from 175,000 people infected with Covid in 2020 and studied their clinical results.
In March and April 2020, nearly one in five patients hospitalized with COVID-19 in the US died from the disease.
In September and October this fell by almost half to 8.6 percent.
The researchers were able to use AI to predict whether a patient would be more likely to have a serious case based on demographic factors, pre-existing conditions, and measurements taken on a patient’s first day in hospital.
Such models could be useful for hospitals that are now filling up again as the Indian ‘Delta’ variant spreads rapidly through undervaccinated parts of the US
Researchers developed models that can predict a Covid patient’s clinical outcomes based on measurements taken on the day they arrive at the hospital. Pictured: A woman talks to her husband at his bedside in a hospital in Fullerton, California
Improvements in care made patients less likely to have severe Covid cases or die from the disease as the pandemic continued
As the pandemic progressed, doctors got better at treating COVID-19 patients.
This pattern can be attributed to the experiences of frontline health professionals as well as ongoing research into the best treatments for patients.
However, such research is more difficult in the US than in other countries because patient records in the US are not standardized for all health systems.
In the UK, for example, all patients in the country entered their health records anonymously into a centralized system. Researchers use this system to identify patterns and improve treatment.
To improve data availability in the US, researchers have put together a similar – albeit smaller – centralized system called the National COVID Cohort Collaboration.
The research – done by a group of scientists from the University of Colorado, Johns Hopkins University and others – was published on Tuesday in JAMA Network Open.
The scientists analyzed data from 175,000 American adults diagnosed with Covid in 2020.
Each patient was categorized according to the severity of their Covid attack, using World Health Organization criteria.
Of those 175,000 patients, about 32,000, or 18.6 percent, went to the hospital. Most of these patients stayed overnight – the median length of stay was five days.
About 6,600 patients — 20.2 percent of the hospitalized group — had a severe clinical outcome.
These outcomes included mechanical ventilation, death and discharge to hospice (indicating a terminally ill patient).
Patients hospitalized in the early months of the pandemic were more likely to die than those who became ill later.
The death rate for patients in this study fell from 16.4 percent in March to April 2020 to 8.6 percent in September to October.
Other studies have found similarly steep declines. a piece of paper published online in October reported that mortality fell from 25.6 percent in March to just 7.6 percent in August.
As more information became available about Covid treatments, doctors became less likely to use some drugs and more likely to use others
The researchers found that treatments used in U.S. hospitals changed over time as more information became available about how well different options worked.
Hydroxychloroquine, for example, fell out of favor, while remdesivir became more and more common. About 17 percent of patients in the study received remdesivir during their hospital stay.
Doctors also became less likely to use invasive equipment such as ventilators and extracorporeal membrane oxygenation (or ECMO) — a machine that provides heart and respiratory support — as the pandemic continued.
When analyzing patient demographics, the researchers found several factors that made patients more likely to have a severe case of Covid.
Of the 32,000 hospitalized patients, 41 percent had at least one pre-existing condition. The most common condition was diabetes – which affected 26 percent of patients.
Other common conditions in the more severe patients were older age, male gender, liver disease, dementia and obesity.
Black and Asian patients were also more likely to have worse clinical outcomes.
In addition, the researchers identified measurements taken early in a patient’s hospital stay that may indicate a worse outcome later.
Those at-risk patients had higher initial blood pressure, oxygen saturation and inflammation, among other things.
As a result, the researchers used AI to develop models — predicting a patient’s clinical outcomes based on information available on their first day in the hospital.
While the researchers caution that more research is needed in this area, these models could provide the basis for tools actively used in hospitals to determine which patients are prioritized.
Such models could be invaluable as the Delta variant is causing new outbreaks in undervaccinated parts of the US, causing some hospitals to fill up again.
Previous research has shown that hospitals have higher death rates when they become overrun with patients. Smarter, data-driven care can help address this problem.