Hospital-Wide Sepsis Detection: A Machine Learning Model Based on Prospectively Expert-Validated Cohort
Krankenhausweite Sepsisdetektion: Machine-Learning-Modell auf Basis prospektiv expert:innen-validierter Kohorten
Journal of Clinical Medicine · 21.01.2026 · 15 (2) · 855
DOI: 10.3390/jcm15020855 ↗Auf einen Blick
Ein krankenhausweites Machine-Learning-Modell zur Sepsisdetektion zeigte eine höhere diagnostische Leistung als konventionelle Scores. Der BiAlert Sepsis model erreichte eine AUC-ROC von 0,95.
Fragestellung
Kann ein krankenhausweites Machine-Learning-Modell Sepsis besser erkennen als konventionelle Scoring-Systeme?
Methode
Retrospektive Kohortenstudie mit 218.715 Krankenhausepisoden (2014–2018); Entwicklung und Evaluation von 30 Modellen, darunter Random Forests und neuronale Netze, mit Trainings-/Testsplit 5/7 zu 2/7.
Ergebnisse
BiAlert Sepsis model: AUC-ROC 0,95, Sensitivität 0,93, Spezifität 0,84. Gegenüber dem besten regelbasierten Verfahren sank die Zahl falscher Positive um 39,6 % (13,10 % vs. 21,70 %; p < 0,01).
Einschränkungen
Externe Validierung und prospektive Studien zu klinischen Auswirkungen werden benötigt.
Schlussfolgerung der Autor:innen
Das krankenhausweite Machine-Learning-Modell zeigte eine bessere Sepsisdetektion als konventionelle Scoring-Systeme.
Kernaussagen
- AUC-ROC des BiAlert Sepsis model: 0,95
- Weniger falsche Positive als Sepsis-2 + qSOFA
- Externe Validierung und klinische Impact-Studien fehlen
Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.
Kurz gefasst
Fachartikel in der Zeitschrift Journal of Clinical Medicine, online erschienen am 21.01.2026. Verfasst von Marcio Borges-Sa und 12 weiteren Autor:innen. Beteiligte Einrichtung: Multidisciplinary Sepsis Unit, Intensive Care Unit, Son Llatzer University Hospital, 07198 Palma de Mallorca, Spain. Die Arbeit wurde bisher 8-mal zitiert (Crossref) und nennt 22 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Instituto de Salud Carlos III, European Regional Development Fund/European Social Fund.
Abstract (Originalsprache)
Background/Objectives: Sepsis detection remains challenging due to clinical heterogeneity and limitations of traditional scoring systems. This study developed and validated a hospital-wide machine learning model for sepsis detection using retrospectively developed data from prospectively expert-validated cases, aiming to improve diagnostic accuracy beyond conventional approaches.
Methods
This retrospective cohort study analysed 218,715 hospital episodes (2014–2018) at a tertiary care centre. Sepsis cases (n = 11,864, 5.42%) were prospectively validated in real-time by a Multidisciplinary Sepsis Unit using modified Sepsis-2 criteria with organ dysfunction. The model integrated structured data (26.95%) and unstructured clinical notes (73.04%) extracted via natural language processing from 2829 variables, selecting 230 relevant predictors. Thirty models including random forests, support vector machines, neural networks, and gradient boosting were developed and evaluated. The dataset was randomly split (5/7 training, 2/7 testing) with preserved patient-level independence.
Results
The BiAlert Sepsis model (random forest + Sepsis-2 ensemble) achieved an AUC-ROC of 0.95, sensitivity of 0.93, and specificity of 0.84, significantly outperforming traditional approaches. Compared to the best rule-based method (Sepsis-2 + qSOFA, AUC-ROC 0.90), BiAlert reduced false positives by 39.6% (13.10% vs. 21.70%, p < 0.01). Novel predictors included eosinopenia and hypoalbuminemia, while traditional variables (MAP, GCS, platelets) showed minimal univariate association. The model received European Medicines Agency approval as a medical device in June 2024.
Conclusions
This hospital-wide machine learning model, trained on prospectively expert-validated cases and integrating extensive NLP-derived features, demonstrates superior sepsis detection performance compared to conventional scoring systems. External validation and prospective clinical impact studies are needed before widespread implementation.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Multidisciplinary Sepsis Unit, Intensive Care Unit, Son Llatzer University Hospital, 07198 Palma de Mallorca, Spain.
- Multidisciplinary Sepsis Group, Health Research Institute of the Balearic Islands (IdISBa), 07198 Palma de Mallorca, Spain.
- Balearic Islands University (UIB), 07198 Palma de Mallorca, Spain.
- Faculty of Medicine, Finis Terrae University, 7501015 Santiago, Chile.
- Critical Care Department, Clinica Las Condes Hospital, 7591046 Santiago, Chile.
- Instituto de Ingenieria del Conocimiento (IIC), 28049 Madrid, Spain.
ORCID-Profile: Marcio Borges-Sa, Andres Giglio, Antonia Socias, Cristina Pruenza, Gonzalo Hernández, Elisa Martin, Ignacio Martin-Loeches
Förderung
- Instituto de Salud Carlos III
- European Regional Development Fund/European Social Fund
Zitieren
APA
Borges-Sa, M., Giglio, A., Aranda, M., Socias, A., Del Castillo, A., Pruenza, C., Hernández, G., Cerdá, S., Socias, L., Estrada, V., de la Rica, R., Martin, E., & Martin-Loeches, I. (2026). Hospital-Wide Sepsis Detection: A Machine Learning Model Based on Prospectively Expert-Validated Cohort. Journal of Clinical Medicine, 15(2), 855. https://doi.org/10.3390/jcm15020855
Vancouver
Borges-Sa M, Giglio A, Aranda M, Socias A, Del Castillo A, Pruenza C, et al. Hospital-Wide Sepsis Detection: A Machine Learning Model Based on Prospectively Expert-Validated Cohort. Journal of Clinical Medicine. 2026;15(2):855. doi:10.3390/jcm15020855
BibTeX
@article{BorgesSa2026,
author = {Borges-Sa, Marcio and Giglio, Andres and Aranda, Maria and Socias, Antonia and Del Castillo, Alberto and Pruenza, Cristina and Hernández, Gonzalo and Cerdá, Sofía and Socias, Lorenzo and Estrada, Victor and de la Rica, Roberto and Martin, Elisa and Martin-Loeches, Ignacio},
title = {Hospital-Wide Sepsis Detection: A Machine Learning Model Based on Prospectively Expert-Validated Cohort},
journal = {Journal of Clinical Medicine},
year = {2026},
volume = {15},
number = {2},
pages = {855},
doi = {10.3390/jcm15020855}
}
EUVIMED