Big Data Sheds New Light On Leading Cause Of Death In Hospitals
Hospitals should reconsider the prevailing one-size-fits-all approach for treating sepsis, according to new research from the University of Pittsburgh.
A quarter of a million Americans die from sepsis each year. The disease occurs when the body responds to a severe infection by damaging its own tissues and organs.
Pitt researchers fed the medical records of thousands of sepsis patients across the country into a machine learning alogirthm, which determined that there are actually four subtypes of sepsis. The subtypes vary in how aggressive they are and how they affect the body.
"There are so many differences in the types of infection and types of patients," said lead author and Pitt professor Christopher Seymour. "Perhaps there are different types of sepsis that could be treated differently."
The types range from Alpha, the most common with the least abnormal lab test results, to Delta, the least common but most deadly type.
For years, according to Seymour, no breakthroughs have been made in treating sepsis. He said that might have been because researchers were lumping all sepsis patients in together.
"When we looked at clinical trials that have already been completed, and often concluded with no benefit, we found that some of these patients actually may have benefited from these treatments depending on which sepsis types they had," Seymour said.
Seymore said researchers should next try to further refine the four sepsis subtypes. They also want to figure out which of the 29 variables the AI used in its analysis are most important for identifying each subtype. Seymour said eventually, they'll use this information to guide clinical trials, to see if certain treatments are more effective on different types of sepsis.