Real People. Real Results.®

A new report suggests sepsis prediction tool misses many cases

On Behalf of | Jul 5, 2021 | Failure to Diagnose

Imagine someone came into your workplace and said they had a new tool that could predict problems before they happen. The salesman tells you it has an accuracy rate of between 73 and 86%. Would you be tempted? Or would you want to see independent evidence to back up those statistics before parting with your money?

Sadly, a recent study suggests that many hospitals took the salesman at their word when it came to a tool claimed to predict “the probability of a likelihood of sepsis.” Researchers at the University of Michigan Medical School put the accuracy rate of the Epic Sepsis Model algorithm at only 63%. They suggest too many medical providers who purchased the tool did not do their homework. If they did, they would have found a lack of independent evidence to back up the manufacturer’s claims.

Why is predicting sepsis so important?

Sepsis is one of the biggest killers in hospitals, so any improvements in prediction or detection are to be applauded. A tool that gets it right 63% of the time is still not bad, provided people understand its level of accuracy. The report suggests that this was not the case.

What is worse, the study reports the algorithm predicted the cases that doctors themselves would spot without it. It was less good at picking up the ones they missed. As a result, it may have led to even more missed cases of sepsis. When an intelligent algorithm that people tell you has a high accuracy rate agrees with you, you are more likely to believe you have spotted all the cases. If someone told you the algorithm only spots 10% of cases, you would assume you need to search harder.

Hospitals and doctors must do all they can to prevent and identify sepsis. Relying on unproven technology is not an excuse for missing a sepsis diagnosis that leads to your suffering. If you’ve been injured by a doctor’s mistake, make sure that you find out more about your right to compensation for your losses. 

FindLaw Network