Photo illustration -Human intelligence provides the purpose, judgment and responsibility. Artificial intelligence expands what humanity can achieve.
An early warning system at UC San Diego helped nurses and physicians recognize a deadly emergency sooner. In a study of more than 6,200 patients with sepsis, its use was associated with a 17% relative reduction in hospital deaths.
By: Ya Libnan – Article 7 in our series about how AI is helping humanity: Analysis
Imagine arriving at an emergency department feeling terribly ill. Your heart is racing. You may have a fever, or perhaps your temperature is unusually low. The staff is working to find out what is wrong, but the danger is moving faster than any one symptom can explain.
It could be sepsis: a life-threatening reaction to infection that can damage organs. The sooner clinicians recognize it and begin appropriate treatment, the better the patient’s chance. Yet its early warning signs can be scattered among vital signs, laboratory results and medical history. In a busy emergency department, putting those pieces together is difficult.
At UC San Diego Health, researchers and medical staff put artificial intelligence to work on that problem. Their system, called COMPOSER, continuously examines more than 150 patient variables, including vital signs, laboratory results, medications and medical history. When it detects a pattern suggesting sepsis, it sends an alert to a nurse. The nurse reviews the patient and can bring the concern to a physician, who decides what care is needed.
That is the partnership I want this series to celebrate. AI can watch for a pattern across thousands of changing details. A nurse can assess the person in front of her. A physician can weigh the alert against everything else the team knows and act. Each contributes something the others cannot.
The result deserves attention. In a study published in npj Digital Medicine, researchers examined 6,217 adults with sepsisat two UC San Diego emergency departments before and after the system was introduced. They found that COMPOSER’s deployment was associated with a 17% relative reduction in deaths in the hospital from sepsis. Expressed another way, the estimated difference was 1.9 percentage points, not 17 percentage points. The study also found better adherence to recommended sepsis care, and antibiotics were administered sooner in cases where nurses escalated an alert to physicians.
Those figures are encouraging, but accuracy matters. This was a before-and-after study, not a randomized trial. It cannot prove that the AI system alone caused every improvement, and its performance should be tested in other hospitals. The researchers themselves called for broader trials. We do not need to exaggerate the finding to recognize its importance: a working hospital system was linked to better care and fewer deaths among people facing a medical emergency.
Think of what this means to a family waiting for news. They do not care whether the first warning came from a computer or from someone who noticed a change at the bedside. They care that the warning reached a trained team in time to help. The technology’s value lies in making that human response more timely.
When people ask whether AI is helping humanity, here is one answer: it is already helping clinicians notice danger that might otherwise be recognized too late. It does not replace the nurse who responds or the doctor who treats the patient. It gives them another chance to be there when every minute matters.
Sources: Original study, npj Digital Medicine (2024); UC San Diego Health study summary.
