Science

This algorithm could save thousands of lives in intensive care – Science et Avenir

In France, a person dies of sepsis every four seconds. However, this syndrome is largely misunderstood. When the body is infected, it sometimes responds with a strong immune response. This defense mechanism sometimes causes sepsis (formerly called “sepsis”), an unregulated body response to severe infection.

The infection can begin locally (peritonitis, pneumonia, urinary tract infections, etc.) and spread to other organs. Then the patient’s condition worsens: inflammation becomes generalized and can lead to the formation of blood clots, vasodilation, organ failure and death. In this case, the situation develops rapidly and the patient’s health deteriorates rapidly. So much so that about 35% of those in intensive care die from it in the United States.

Every year, 70,000 French people get sepsis and 30,000 die from it. PFor survivors, the consequences can be severe, such as amputations, damage to the kidneys or lungs.

One in five lives saved

For the treatment of sepsis, patients are prescribed broad-spectrum antibiotics that are effective against a large number of bacteria. The sooner they are introduced, when the sepsis has not yet gone too far, the greater the chance of recovery. However, sepsis is very difficult to recognize among other symptoms of the patient.

It is characterized by fever and confusion. “Severe sepsis is easy to recognize because organ damage is obvious. But at this point, it becomes very difficult to correct the reaction of the immune system and stop it, laments Dr. Martin Doerfler from the Northwell Health Training Center. and Innovation from Sciences et Avenir. On the other hand, the first stages of sepsis are confused with a whole host of diseases and even the body’s response to a simple surgical intervention.

“Apart from saving time by even earlier detection of sepsis, there is no other way to improve the survival rate of these patients,” explains Professor Suchi Saria, machine learning specialist for the medical world at Johns Hopkins University. So to try to detect sepsis further downstream, an American university has developed an algorithm capable of identifying at-risk patients. The program, called the Targeted Real-Time Early Waning System, analyzes the patient’s medical history, current symptoms, and lab results.

If a patient shows symptoms of sepsis, the doctors in the ward raise an alarm. The doctor comes to the patient’s bedside, where he must fill out a short questionnaire about the patient’s condition. The attending physician must then decide whether to administer antibiotics to the patient.

The algorithm, tested on more than 700,000 patients in a two-year clinical trial, was able to detect sepsis an average of 1.85 hours earlier than in units without the algorithm, helping to treat patients much sooner. Compared to cases where the patient died from sepsis, the algorithm was even able to detect sepsis an average of six hours earlier than with traditional care. As a result, the death rate among patients decreased by 18.2%, the number of lives saved, Dr. Sariya welcomes. “With sepsis, every hour counts. This is a huge leap forward that will save thousands of lives every year.” The results of the study were published in the journal Nature Medicine.

A drug widely used in intensive care

The algorithm performed well with a sensitivity level of 82% (meaning that 82% of the time the alarm went off, the patient actually had sepsis). Moreover, upon arrival, the doctors were able to directly confirm that it was indeed sepsis in 38% of cases.

The authors conducted a second study to see if the service’s medical personnel accepted this new type of instrument or not. Result: 89% of healthcare professionals have accepted the algorithm well in the intensive care unit. A very encouraging figure, because artificial intelligence is currently almost not used in intensive care in hospitals. In medicine, it is mainly used to better understand diseases such as cancer or rare diseases, and to better classify subgroups of patients. In view of these positive results, the team will now attempt to adapt this technology to other common critical care pathologies such as pressure sores or respiratory distress syndromes.

Back to top button

Adblock Detected

Please consider supporting us by disabling your ad blocker.