Loading ...

Anand Viswanathan, MD, had a sinking suspicion he couldn’t quite explain. His patient had unexplained abdominal pain a few days after getting a feeding tube. Imaging showed nothing and vital signs were normal. Then, while Viswanathan was still trying to decipher the problem, the patient’s white blood cell count began to plummet.

None of the signs were quite right, but Viswanathan — a hospitalist at NYU Langone Health, New York City, and medical director of the Inpatient Surgical Unit — started antibiotics targeting an intra-abdominal infection. He had a hunch the patient was septic.

“Pain and dropping white blood count doesn’t fit sepsis,” Viswanathan said. In fact, you’d expect the white blood cell count to skyrocket with infection. “But it’s been ingrained in every hospitalist that sepsis should always be in our minds as something going on.” 

It creates a sort of “sixth sense,” he said.

Antibiotics were the right move, “even if it wasn’t obvious at the time,” Viswanathan said. That night the patient was transferred to the intensive care unit and sepsis was confirmed. Unfortunately, the patient passed away the next day. “Even though I’m not sure, I wonder if starting antibiotics delayed deterioration by a few hours,” Viswanathan said.

Every day, hospitalists across the United States are on the frontlines of managing sepsis. The elusive syndrome is quick and unpredictable, challenging providers to make a complicated diagnosis in a short window of time. While some progress has been made, experts say that existing guidelines fall short and there’s a serious need for more data — especially for sepsis cases that start in-hospital. Sepsis management leaders are looking for specialized teams, more tailored treatment, and maybe artificial intelligence (AI) to curb the insidious syndromes’ impact.

“No question [sepsis is] a huge challenge,” said Chanu Rhee, MD, MPH, director of the Center for Sepsis Epidemiology and Prevention Studies at Harvard Pilgrim Health Care Institute, Boston. It contributes to 1.7 million US hospitalizations each year and 270,000 deaths, making it one of the country’s leading causes of hospital deaths. These cases can progress from mild to life-threatening in just hours.

“Hospitalists often don’t get enough attention for the important role they have in sepsis detection and treatment,” Rhee said.

The vast majority of sepsis cases start outside the hospital and are caught in the emergency department, but around half of those patients go on to be managed in the hospital ward. The remaining 10%-15% of sepsis cases develop in the hospital. Rhee said hospitalists are often the ones responsible for identifying these — most dangerous — cases.

According to the literature, timely infection treatment is the cornerstone of sepsis treatment and a key predictor of mortality. The problem is that the diagnosis and treatment selection are not always straightforward. A 2021 review of sepsis cases identified more than 10,000 signs and symptoms. A patient’s complaint may be as nebulous as “I don’t feel good,” said David Page, MD, MSPH, director of the Medical Critical Care Unit at The University of Alabama at Birmingham and physician advisor for inpatient sepsis.

Even the classic sepsis symptoms, like fever and tachycardia, mimic other conditions that can cause diagnosis delays.

“There’s no pathological gold standard, there’s no test that can prove or refute the diagnosis. It’s really a constellation,” Rhee said.

Clinicians have only hours to parse through lactate tests, blood cultures, and liver function tests to diagnose sepsis and pinpoint the underlying pathogen — of which there are many. And even then, “only 20% of sepsis cases have a positive blood culture,” Rhee said.

And almost half of patients with sepsis don’t have a known pathogen.

Redefining Good Sepsis Care 

In 2015, to expedite sepsis treatment, the Centers for Medicare and Medicaid Services launched sepsis care bundles known as Severe Sepsis and Septic Shock Management Bundle or SEP-1. The bundles mandate a list of actions — lactate test, blood culture, fluids, etc. — to be done in the first 3 and 6 hours of sepsis diagnosis.

While the bundles have helped expedite care, many experts believe it’s time to think beyond them.

“Not all of the mandated interventions are beneficial, and some may actually be harmful,” like intravenous fluid administration, Page said. Focusing on bundle compliance uses resources for interventions that data suggest aren’t helpful to patients.

In a 2021 study, Rhee’s group showed that SEP-1 implementation was associated with an increase in lactate testing, but no change in short term mortality. While the bundles offer a good building block for sepsis care, “it’s time to move to a more refined focus on sepsis outcomes as a quality measurement,” Rhee said, not just proper actions in the first few hours of care.

Plus, most of the data for these bundles were based on emergency department patients and may not transfer well to patients who acquired sepsis in-hospital, Matthew Dettmer, MD, a critical care physician who coordinates in-hospital sepsis management for Cleveland Clinic, Cleveland, and lead author of the paper, said. For example, most hospital patients are already on fluid rendering the fluids mandated by the care bundle unnecessary.

The bundles have served a purpose, Dettmer said.

“There have been significant improvements especially with regard to community-acquired sepsis,” he said. “If we hadn’t made those advancements, I’m not sure we could have exposed the needs of in hospital-onset and put more attention on it.”

Two Kinds of Sepsis

This issue with hospital-onset sepsis is that it’s far more dangerous. Though it impacts a minority of patients (10%-15%), the mortality rate is two-fold higher.

The reason for the disparity is unclear because there’s just not enough data, according to Dettmer. To date, nearly all sepsis treatment guidelines are based on data from patients who became septic outside the hospital — patients who were otherwise healthy.

It could be, Dettmer said, that patients in the hospital are just sicker to begin with or they’re exposed to different pathogens with worse resistance patterns. Hospital-acquired sepsis is also treated outside the highly resourced emergency department. Even in premiere hospital wards, vital signs are checked less frequently, nursing ratios are lower.

“There is research showing there can be delays because they aren’t set up for rapid addressing and treatment,” Rhee said.

Whatever the constellation of drivers, hospital-acquired sepsis has different outcomes, occurs in a different patient cohort, and is diagnosed in a different environment. So in a 2024 paper, Dettmer and coauthors from two other institutions argued hospital-acquired sepsis needed its own research and guidelines.

In fact, because the types of hospital patient who experiences sepsis are so heterogenous, Dettmer said there’s a need for condition-specific guidelines. For instance, the warning signs of sepsis in a patient with cancer with chemotherapy would look very different than a patient with kidney failure. Hospitalists need this condition-specific insight to better support patients, but there’s not enough data yet.

Specialized Teams and Tech

While more data and refined guidelines are still in-process, most hospital systems are already deploying specialized sepsis teams and committees. Ideally, these teams work like a rapid response team, Dettmer said. They’re an independent and fast-acting set of extra hands that also have time to step back and consider the patient more comprehensively, Dettmer said.

Many electronic medical record systems also have some sepsis detection support — notifying doctors when vital signs change — but it’s not quite fast enough, Dettmer said. And they may not be sensitive enough to register changes in already-sick hospital patients. This is where AI could be very valuable.

Several hospital systems have already built and/or deployed AI-based tools for sepsis management. The hope is that AI could pick up on more nuanced changes or more subtle presentations of sepsis and alert hospitalists earlier. These insights could also help better tailor the treatment options and help reduce broad-spectrum antibiotics.

But Rhee said these tools are still controversial. The field needs more data on just how beneficial these AI tools are. “Sepsis diagnosis is complicated, and these tools aren’t that accurate yet. There’s concern about alert fatigue,” he said. “I’m personally not convinced yet.”