NAVOY Sepsis at Skane University Hospital
NAVOY Sepsis predicts sepsis onset three hours in advance from four hours of routinely collected vital parameters, blood gas and laboratory values in ICU patients.
- What this means:
- The source has been read and our reading of it confirmed — the classifications that let this record be compared with others, the AI product, and the clinical use. That is a check of our work, not a judgement on the AI.
- What happens next:
- Nothing, unless this use case is taken up for assessment. Then the AI product and its clinical use are examined together in seven sections — purpose, evidence, CE conformity, legal duties, ethics, safety, and what a hospital must have in place — each signed by a named institution.
- Checked by:
- A named reviewer at the AIH Lab. This is an independent check: neither the organisation nor the vendor did it.
- Submitted
- 2Curated
- 3Maturing
- 4Validated
- 5Scaled
- Clinical problem
- Predict a future risk
- Point of care
- Monitoring
- Nature of AI output
- An alert
- Clinical specialty
- Intensive care
- Care setting
- Intensive care unit
- Patient population
- 304 adult ICU patients (Dec 2020 - Sep 2021), including Covid-19 patients
- Intended use
- NAVOY Sepsis predicts sepsis onset three hours in advance from four hours of routinely collected vital parameters, blood gas and laboratory values in ICU patients.
Full study description
EUROPEAN and RANDOMISED — the strongest study design in this batch. Journal of Critical Care 2024 (PMID 38245375): a prospective randomised clinical validation at the ICU of Skane University Hospital Malmo, with a silent-mode arm. Authors report the algorithm "could predict the development of sepsis, according to the Sepsis-3 criteria, three hours before sepsis onset with high performance: accuracy 0.79; sensitivity 0.80; and specificity 0.78." IMPORTANT CAVEAT: the trial validates PROGNOSTIC ACCURACY only — it reports no patient-outcome difference (mortality, length of stay, antibiotic timing) between arms. Single centre, n=304. First and last authors are affiliated with the vendor AlgoDx AB. Recorded as pilot: provisional, time-boxed research use rather than routine service.
- AI technique
- Classical machine learning
- Input data
- Structured EHR data, Vital signs, Laboratory results
- Output type
- Alert
- Autonomy level
- Informs a human (advisory)
- Model provenance
- Vendor proprietary
- Status
- Pilot
- Country
- Sweden
- Data identifiability
- Not stated in the source
Similar deployments
AIEvidence records
Studies and evaluations attached to this use case.
Contributors
- Deploying organisation
- Skåne University Hospital · Hospital / health system · Sweden
- AI vendor
- AlgoDx AB · Sweden
- Product name
- NAVOY Sepsis
- Source record
- pubmed.ncbi.nlm.nih.gov · PMID:38245375