Photo Credit: iStock.com/Nemes Laszlo

The performance of an antimicrobial resistance risk stratification model for patients with sepsis correlated with local rates of antimicrobial resistance and varied widely across 10 Midwest hospitals and patient subgroups, according to a study published in JAMA Network Open.

“This variability suggests caution is needed when using generalized models for predicting resistant GNB [gram-negative bacilli] etiologies in sepsis,” wrote corresponding author Maria Cristina Vazquez Guillamet, MD, of the Washington University School of Medicine, St. Louis, Missouri, and study coauthors.

Current sepsis treatment guidelines use prediction models to define comorbidities associated with an increased risk of resistant GNB, while also noting the importance of local rates of resistance. Nevertheless, the models do not perform well across hospitals.

In this study, researchers investigated whether patient case mix and local GNB resistance rates contributed to the variable performance of a general risk stratification model for community-onset and hospital-onset sepsis across hospitals. The retrospective analysis included 39,893 adult patients with sepsis at rural and urban hospitals in the greater St. Louis region of Missouri and Illinois. Patients in the study had blood cultures indicating sepsis, received 4 days of antibiotic treatment, and had organ dysfunction.

Researchers stratified culture results for ceftriaxone-susceptible GNB, ceftriaxone-resistant but cefepime-susceptible GNB (RS), and ceftriaxone- and cefepime-resistant GNB (RR). They also developed separate deep learning models for community-onset and hospital-onset sepsis to project levels of antimicrobial resistance in GNB and tested them across hospitals and patient subgroups.

RS contributed to 3.9% of community-onset and 5.7% of hospital-onset sepsis episodes, according to the study. RR contributed to 1.8% of community-onset and 3.9% of hospital-onset sepsis episodes.

“The data show that the risk for resistant GNB in sepsis differs significantly between hospital- and community-onset sepsis and across hospitals,” researchers reported. “The risk appears to depend on the RS and RR prevalence rates in hospitals and subgroups of patients and does not solely depend on patient comorbidities.”

The history of several infections and previous exposure to antibiotics were the most relevant factors associated with ceftriaxone- and cefepime-resistant GNB, the study found.

“Future efforts to benchmark antimicrobial use may need to include community and hospital resistance rates (warranting availability of microbiology data) to accompany patient-level variables,” researchers wrote.