Skip to content

SEPSI Score

PREVIA MEDICAL has developed a digital biomarker SEPSI SCORE, which allow early detection of patients at risk of developing sepsis.

Illustrations Previa 14
Biomarqueurs Sans Fond 4

The sepsis

Sepsis is a generalised inflammatory response associated with infection. If it occurs after an invasive procedure, it is known as a nosocomial infection. Sepsis can be caused by bacteria or certain viruses (SARS-Cov2, H1N1 influenza, etc.).

Sepsis is currently responsible for 60,000 deaths a year in France and 1 in 5 deaths worldwide. Each hour's delay in diagnosis reduces the patient's chances of survival by 7%.

With a 30-day mortality rate of 20-30%*, sepsis is one of the leading causes of death in hospitals worldwide.
A precise algorithm and real-time alerts can significantly reduce the length of stay and mortality rate.

Construction of Digital Biomarkers

Digital biomarkers make it possible to collect precise, real-time information about an individual’s state of health. This data is then transmitted and processed electronically, enabling continuous health monitoring, early detection of changes and prediction of health problems.

Objective of the digital biomarker
First, clearly define the digital biomarker’s objective: the disease or medical condition it is intended to predict
1
Data collection
We gather data relevant for score caluclation. This includes clinical data: Vital signs, laboratory results, medication, patient history
2
Data pre-processing
Before analysis, data is typically pre-processed to eliminate errors and outliers
3
Features selection​
We identify the relevant characteristics or variables in the data that can be used to construct the biomarker
4
Model training​
We use a training dataset to validate our model
5
Biomarker performance
Performance monitoring (Sensitivity, Specificity, Precision) is conducted through a quality approach using post- market surveillance. See retrospective study (Thiboud et al., 2025)
6
Clinical validation
Before deploying our digital biomarker in a clinical setting, we validated it through clinical studies to confirm its effectiveness and reliability in routine practice
7

Training and validation of our algorithm

2 Mds
45 127
1 308
5 270
139

Data analysed

Training

Stays

Training

Sepsis

Detected in training

Stays

Score validation

Sepsis

Validated by an infectiologist

Performance of the SEPSI score in prospective use

The ability to identify patients at risk of developing sepsis before the first symptoms appear

Sensitivity
0 %
Specificity
0 %
Positive predictive value
0 %

Publication Scientific article

The SEPSI Score was developed and validated in collaboration with a hospital centre over the course of a 22-month retrospective study. This R&D work led to the publication of an article in the journal Diagnostics in January 2025, entitled ‘Development and Validation of a Machine Learning Model for Early Prediction of Sepsis Onset in Hospital Inpatients from All Departments’.

The study validated our model with a positive predictive value of 0.610, significantly outperforming early warning scores such as SIRS, MEWS and qSOFA. In addition, the model demonstrated high sensitivity and specificity (0.845 and 0.987 respectively), making it possible to accurately predict the risk of developing sepsis up to 48 hours* before the first clinical signs.

Results of the retrospective study

Sensitivity
0 %
Specificity
0 %
Positive predictive value
0 %
 

SEPSI Score

Integrated into the hospital’s IT workflows, SEPSI SCORE’s purpose is to support earlier clinical assessment by providing areal-time risk score computed from routinely collected electronic health record (EHR) data. 

Previa Pictos 01

Early detection of patients being at risk of sepsis

Previa Pictos 03

Alerting healthcare professionals

Previa Pictos 04

Risk score calculated every hour

Previa Pictos 05

explainability information supporting clinician interpretation

The PREVIA © platform

Discover the features of our solution