An Algorithm That Predicts Fatal Infections Is Often Failing

A complication of infection known as sepsis is a digital assassin in American hospitals. So it’s no surprise that more than 100 health systems use an initial warning system offered by Epic Systems, the dominant provider of U.S. electronic health records. The system launches alerts based on a proprietary formula that instantly watches the signs of the situation in the results of a patient’s tests.

But a new study using data from nearly 30,000 patients at University of Michigan hospitals suggests that the Epic system is malfunctioning. The authors say he missed two-thirds of cases of sepsis, rarely found cases that the medical staff did not notice, and often sounded false alarms.

Karandeep Singh, an assistant professor at the University of Michigan who led the study, says the results illustrate a broader problem with the owner. algorithms increasingly used in health care. “They’re very popular, and even there’s little published on these models,” Singh says. “It’s shocking to me.”

The study was published Monday in Internal medicine JAMA. An Epic spokesman disputed the study’s findings, saying the company’s system had “helped clinicians save thousands of lives.”

Epic’s is not the first widely used health algorithm to raise concerns that the technology supposed to improve health care is not providing, or even actively harming. In 2019, a system used on millions of patients to prioritize access to special care for people with complex needs was found. lowball the needs of black patients compared to white patients. What he ordered some Democratic senators asking federal regulators to investigate the bias in health algorithms. A study published in April found that statistical models used to predict suicide risk in mental health patients performed well for white and Asian patients but poorly for black patients.

The way sepsis attacks hospital quarters has made it a particular target of algorithmic aids for medical staff. Guidelines from the Centers for Disease Control and Prevention to health care providers in sepsis encourage the use of electronic medical records for surveillance and prevention. Epic has several competitors that offer commercial warning systems, and some American research hospitals have he built his instruments.

Singh says automated sepsis warnings have enormous potential because key symptoms of the condition, such as low blood pressure, can have other causes, making it difficult for staff to detect them early. Start treatment of sepsis as antibiotics only an hour before it can make a big difference to patient survival. Hospital administrators often take a particular interest in the sepsis response, in part because it contributes to U.S. government hospital assessments.

Singh runs a lab in Michigan researching applications automatic learning to the care of patients. He is curious about Epic’s sepsis warning system after being asked to chair a committee on the university’s health system set up to oversee the use of machine learning.

When Singh learned more about the tools in use in Michigan and other health systems, he became concerned that they came mostly from vendors who had little information about how they worked or performed. Its own system had a license to use Epic’s sepsis prediction model, which the company told customers was highly accurate. But there had been no independent validation of their benefits.

Colleagues from Singh and Michigan tested Epic’s forecasting model on records for nearly 30,000 patients covering nearly 40,000 hospitalizations in 2018 and 2019. Researchers have noted how often Epic’s algorithm has marked the people who have developed it. sepsis as defined by the CDC and the Centers for Medicare and Medicaid Services. And they compared the alerts that the system would trigger with staff-registered sepsis treatments, which did not see Epic sepsis alerts for patients included in the study.

The researchers say their results suggest that the Epic system would not make a hospital much better at catching sepsis and could burden staff with unnecessary warnings. The company’s algorithm did not identify two-thirds of the 2,500 cases of sepsis in Michigan data. It would be recommended for 183 patients who developed sepsis but who had not received timely treatment from staff.

Source link

Read More

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button