Blood test identifies sepsis pathogen in hours, not days

Clinical takeaway: A same-day organism ID plus susceptibility call could let clinicians target antibiotic therapy inside the treatment window and de-escalate off broad-spectrum drugs sooner. 

When a patient is septic, an antibiotic is given long before the lab knows what it is treating. Blood culture, the standard for identifying the organism behind a bloodstream infection, takes two to seven days to return a definitive answer with susceptibility testing adding even more time. So, clinicians start broad-spectrum options and then wait. 

Faster alternative diagnostics exist, but each represents tradeoffs. Culture-free methods that pull pathogen DNA straight from blood can move quickly, yet they falter at the very low bacterial loads many real infections carry, and they do not return the phenotypic susceptibility a clinician needs to choose a drug. Other direct-isolation approaches hit a sensitivity floor around five bacteria per milliliter, above which a meaningful share of cases actually sit. 

The open question has been whether one workflow could keep culture's sensitivity and breadth while collapsing its timeline, and deliver both the organism and its drug susceptibility. Researchers now report a proof-of-concept platform built to do exactly that, pairing a rapid enrichment step with single-cell identification and susceptibility testing. 

The platform delivered a full answer, the organism and the antibiotics it would respond to, in as little as 6.75 hours, against the two to seven days blood culture takes. Speed varied by target: harder organisms like Staphylococcus aureus ran longer, out to 17 hours, and the average across cases was about nine. 

Accuracy tracked the hospital lab closely. On more than 100 positive cultures, the platform named the organism with 96.15% agreement in one to two and a half hours, and misidentified none. Four samples resolved only partway, to family or Gram level. 

On susceptibility, across 219 drug and dose combinations, it landed the right category, susceptible, intermediate, or resistant, 93.9% of the time, and never called a susceptible organism resistant. But 4.1% of resistant organisms were called susceptible, the dangerous direction: it steers a physician toward a drug that will not work. 

It also worked at very low bacterial loads, down to 0.1 to 1 organism per milliliter, below where competing culture-free methods lose reliability. 

The team validated the platform in two ways. For identification, they ran deidentified positive blood cultures from Penn State's Hershey Medical Center microbiology lab and compared results against the lab's own reports. For speed, sensitivity, and susceptibility, they used donor whole blood spiked with known clinical isolates at controlled concentrations. The enrichment step grows and separates bacteria out of whole blood while red cells settle, and single-cell imaging with molecular barcoding then names the organism and reads its drug response from a handful of cells. 

What stands between this and the bedside is real-world validation in patients. The results so far come from stored cultures and spiked blood, not from prospective testing on patients whose infections are still unknown, and the authors point to multi-center studies as the next step.  
 
Two practical hurdles remain: prior antibiotics, which many septic patients receive before blood is drawn and which can suppress the growth the method depends on, and polymicrobial infections, where a fast-growing organism can mask a second one. The team also flags cost and workflow complexity, and points to automation and AI-assisted imaging as the path to making the platform practical in a working lab. 

"We are developing a comprehensive diagnostic platform that rapidly tells physicians both the specific bacteria causing a bloodstream infection, as well as the ideal antibiotic to treat the infection, before sepsis ever sets in," said Pak Kin Wong, PhD, professor of biomedical engineering, Pennsylvania State University. 

Source: Chin SM, et al. (2026 Aug 26) Sci Adv. Rapid and Robust Diagnosis of Bloodstream Infections by Single Cell Analysis