Sepsis ImmunoScore

Performance Metrics

Clinical Evidence

Two full peer-reviewed papers anchor the evidence base, both with manufacturer involvement. The development-and-validation study in NEJM AI (2024) enrolled adults with suspected infection prospectively at five US institutions between April 2017 and July 2022 — 2,366 patients for algorithm development, 393 for internal validation and 698 for external validation — and reported an area under the curve of 0.81 for sepsis within 24 hours in the external cohort. The four risk categories behaved the way a bedside tool needs them to: category-specific likelihood ratios for sepsis ran 0.1 (low), 0.5 (medium), 2.1 (high) and 8.3 (very high), and external-cohort in-hospital mortality climbed from 0.0 per cent in the low band through 1.9 and 8.7 per cent to 18.2 per cent in the very-high band, with the categories also tracking length of stay, ICU utilisation, mechanical ventilation and vasopressor use. The second paper, in Diagnostics (2026), is the head-to-head test a procurement committee will actually want: a retrospective analysis of prospectively collected blood samples and linked EMR data from 6,027 adults with suspected infection across seven US hospitals. For sepsis diagnosis the ImmunoScore reached an AUC of 0.82 against SOFA at 0.72, procalcitonin at 0.70, NEWS at 0.69, qSOFA at 0.67, C-reactive protein at 0.61 and SIRS at 0.59, with similar advantages for predicting in-hospital mortality and ICU admission. The comparison is against scores and single biomarkers, not against a clinician's overall judgment, and the study was funded in part by Prenosis alongside US federal agencies. What is absent from the record so far, and worth being plain about, is any randomised or controlled study showing that deploying the score changes antibiotic timing, ICU utilisation or mortality — the classic gap between a better number and better care.

Inside the Auris+ Listing

Regulatory Approvals

Safety Record

No recalls, FDA safety communications or published adverse-event reports attributable to the Sepsis ImmunoScore were identified in public sources as of August 2026 — read as none found rather than a completed audit, since the FDA MAUDE and recall databases could not be queried directly from this environment. The clinically relevant cautions are structural. The score is an adjunct: a low band does not exclude sepsis and a very-high band does not diagnose it, and the authorization language keeps the treating clinician responsible for the decision. The workflow requires procalcitonin and C-reactive protein results, so the score arrives with laboratory turnaround, not instantaneously at triage. And EHR-embedded sepsis alerting as a class has a history of high false-alarm burden; the ImmunoScore's calibrated four-band design is a response to that history, but its real-world alert burden outside study conditions has not yet been independently published.

Intended Use & Indications

The Sepsis ImmunoScore is artificial-intelligence/machine-learning software as a medical device intended to aid in the risk assessment for presence of, or progression to, sepsis within 24 hours of patient assessment for adults (18 and over) with suspected infection in the emergency department or hospital. The software draws up to 22 predetermined parameters — patient demographics, vital signs, routinely accepted clinical laboratory tests, and the sepsis-specific biomarkers procalcitonin and C-reactive protein — and returns a composite risk score with one of four risk-stratification categories. It is intended to be used in conjunction with other laboratory findings and clinical assessment; it does not diagnose sepsis on its own, and the treating clinician remains responsible for interpretation and patient management.