Proprietary artificial intelligence software designed for early warning systems. Sepsis According to a new study from the University of Michigan, they cannot differentiate between high- and low-risk patients before seeking treatment.

The tool, called the Epic Sepsis Model, is part of Epic’s electronic medical record software, which serves 54 percent of patients in the United States and 2.5 percent internationally, according to a statement from the company’s CEO. Reported From the Wisconsin State Journal. It automatically generates sepsis risk estimates every 20 minutes in hospital records, which clinicians hope will allow them to detect when a patient is at risk before things get worse. Sepsis may occur.

“Sepsis has all these vague symptoms, so when a patient shows up with an infection, it can be really hard to know who can be sent home with some antibiotics and who needs to stay in the intensive care unit. may be needed. We still miss a lot of sepsis patients,” said Tom ValleyAssociate Professor in Pulmonary and Critical Care Medicine, ICU Clinician and co-author of a study recently published in the New England Journal of Medicine AI.

Sepsis is responsible for one-third of all hospital deaths in the United States, and early treatment is key to patient survival. The hope is that AI predictions can help do that, but for now, they don’t seem to be benefiting from patient data as much as clinicians are.

“We suspect that some of the health data that the Epic Sepsis Model relies on encode, perhaps unintentionally, clinicians’ suspicions that a patient has sepsis,” he said. Gina WaynesAssociate Professor of Computer Science and Engineering and corresponding author of the study.

Patients may not receive blood culture tests and antibiotic treatment until they develop symptoms of sepsis, for example. While such data could help AI accurately identify sepsis risks, it may enter the medical record too late to help clinicians advance treatment.

This mismatch in timing between when information is available to AI and when it is most relevant to clinicians was evident in a review by researchers of 77,000 hospitalized patients at the University of Michigan Health, the clinical arm of Michigan Medicine. How did the epic sepsis model for adults perform? .

The AI ​​had already estimated each patient’s risk of developing sepsis in standard medical center operations, so all the researchers had to do was pull the data and do their own analysis. About 5% of patients had sepsis.

To measure the AI’s performance, the team estimated the probability that the AI ​​assigned higher risk scores to patients diagnosed with sepsis, compared to patients who were not diagnosed with sepsis.

When incorporating predictions made by AI at all stages of a patient’s hospital stay, AI can correctly identify a high-risk patient 87% of the time. However, the AI ​​was only accurate 62% of the time when using patient data recorded before the patient met criteria for sepsis. Perhaps most telling, the model assigned high-risk scores to only 53% of patients who developed sepsis when predictors were limited to before blood cultures were ordered.

The results showed that the model was predictive of whether patients had received a diagnostic test or treatment when making predictions. At this point, clinicians already suspect their patients have sepsis, so AI predictions are unlikely to make a difference.

Donna Tjandra, a doctoral student in computer science and engineering and co-author of the study, said, “We need to consider when evaluating a model in a clinical workflow when deciding whether it is appropriate for clinicians. It’s helpful.” “Evaluating the model with data collected after the clinician already suspects the onset of sepsis may strengthen the model’s performance, but it may not be consistent with what clinicians do in practice. There will be help.”

Source: University of Michigan