Sepsis is deadly. According to the World Health Organization, sepsis infections are responsible for about one in five of all deaths worldwide, and most of those deaths are in children. Early treatment is the best option, yet diagnosis can take up to a week. A new study demonstrates how that time can be reduced to just hours (Sci. Adv. 2026, DOI: 10.1126/sciadv.aeh3580).
Very low pathogen load, a wide variety of sepsis-causing microbes, and generic symptoms (fever, clammy skin, confusion, pain) make diagnosing sepsis difficult. For conventional diagnosis, clinicians use blood culture to increase the pathogen concentration to a detectable level. Further subcultures identify the pathogen, then test which drug would be effective against it. This takes days.
But time is a problem when trying to diagnose and treat a fatal infection, says Pak Kin Wong, a biomedical engineer at Pennsylvania State University who led the new work. There are a lot of drug-resistant pathogens, he says, “so it’s very critical to have a quick way to evaluate [which] antibiotics will be effective” and to prevent unnecessary antibiotic use, which contributes to antimicrobial resistance.
Wong’s solution is a method called sedimentation-assisted tandem rocking and enrichment for analysis and monitoring (STREAM), which whittles diagnosis time down to under 9 hours. In this approach, rather than waiting for the bacteria to multiply, samples are collected at multiple points during the culture process and analyzed.
First, the bacteria are separated from the blood cells, while simultaneously enrichedby mixing the blood sample i with a customized “broth” and rocking it gently. Red blood cells clump to the bottom, leaving any bacteria suspended in solution.
Pak Kin Wong at Penn State researches pathogen and characterization and personalized immunotherapy. Credit:
Kate Myers / Penn State
Next, single-cell detection technologies identify the pathogen, and check for antimicrobial susceptibilities. For this, the researchers use sequential fluorescence in situ hybridization (seqFISH), which identifies the bacteria with a color-coded molecular barcode scheme. “We are able to detect a whole panel of common sepsis-causing bacteria using very few number of cells,” Wong says.
For antimicrobial susceptibility testing (AST), the researchers used an imaging technique which encapsulates bacterial cells in gel in a microwell array. They then use the array to monitor the responses of individual cells to antibiotics.
The team tested their approach with a panel of 21 bacteria associated with sepsis and 104 positive bloodstream infection samples. The new diagnostic method achieved over 96% concordance with lab reports, and AST revealed 94% agreement across 219 drug-dose combinations.
The team still has work to do, Wong says, including reducing the 4% error rate in pathogen ID. “We need a much larger panel in order to truly evaluate the accuracy of this approach,” he says. They would also like to expand the panel of sepsis-causing pathogens beyond just bacteria, such as fungal or viral ones.
There are other emerging rapid-diagnostic technologies for sepsis. For example, the SepSeeQ method, based on work by Mads Albertsen’s group at Aalborg University, isolates and sequences microbial DNA and runs it through a classification software.
Wong says the real challenge now lies in making his testing process fast, simple, and robust enough for use in clinical settings. For this, the researchers are looking at automation, by integrating AI and other microfluidic technologies into the workflow.
