TY  - JOUR
AU  - Pylarinou, Charalampia
AU  - Gortzis, Lefteris
AU  - Koletsis, Efstratios
AU  - Leivaditis, Vasileios
AU  - Mavrilas, Dimosthenis
PY  - 2026
DA  - 2026/08/11
TI  - DETER: A Clinical Deterioration Prediction Algorithm to Improve Patient Care with Devices-Based Telemetry and Generative AI
JO  - Journal of Computational Intelligence in Biomedicine
T2  - Journal of Computational Intelligence in Biomedicine
JF  - Journal of Computational Intelligence in Biomedicine
VL  - 1
IS  - 1
SP  - 24
EP  - 43
DO  - 10.62762/JCIB.2025.132871
UR  - https://www.icck.org/article/abs/JCIB.2025.132871
KW  - Generative AI
KW  - clinical deterioration
KW  - machine learning
KW  - remote patient monitoring
AB  - Traditional Early Warning Systems (EWS) relying on intermittent vital signs and aggregate scoring often miss subtle pre-deterioration changes. Recent advances in continuous telemetry and artificial intelligence present opportunities for enhanced detection. We propose DETER, a novel algorithm integrating continuous vital sign telemetry with Retrieval-Augmented Generation (RAG)-enhanced generative AI. The system processes real-time physiological data (ECG, SpO2, blood pressure, temperature, respiratory rate, heart rate variability) alongside electronic medical records and clinical assessments. Developed and validated on 94 patients (74% aged >65 years) from the Cardio-Thoracic Clinic of University Hospital of Patras, with synthetic data augmentation generated from the MIMIC-IV-ED cohort, the Transformer-based architecture achieves 96.5% accuracy with Area Under the Curve (AUC) of 0.98 (6h) and 0.96 (24h). DETER outperforms traditional EWS by providing risk stratification with lead times of days to weeks, not hours, and generates continuously updated, personalized deterioration risk scores for proactive intervention and optimized resource allocation. The integration of continuous telemetry with RAG-enhanced generative AI moves monitoring from reactive to predictive and from population-level to personalised. By forecasting physiological trajectories with lead time in days rather than hours, DETER can enable earlier intervention, reduce preventable deterioration, and support both acute and ambulatory care. Prospective validation is warranted to confirm real-world impact and feasibility in diverse healthcare environments.
SN  - request pending
PB  - Institute of Central Computation and Knowledge
LA  - English
ER  - 

@article{Pylarinou2026DETER,
  author = {Charalampia Pylarinou and Lefteris Gortzis and Efstratios Koletsis and Vasileios Leivaditis and Dimosthenis Mavrilas},
  title = {DETER: A Clinical Deterioration Prediction Algorithm to Improve Patient Care with Devices-Based Telemetry and Generative AI},
  journal = {Journal of Computational Intelligence in Biomedicine},
  year = {2026},
  volume = {1},
  number = {1},
  pages = {24-43},
  doi = {10.62762/JCIB.2025.132871},
  url = {https://www.icck.org/article/abs/JCIB.2025.132871},
  abstract = {Traditional Early Warning Systems (EWS) relying on intermittent vital signs and aggregate scoring often miss subtle pre-deterioration changes. Recent advances in continuous telemetry and artificial intelligence present opportunities for enhanced detection. We propose DETER, a novel algorithm integrating continuous vital sign telemetry with Retrieval-Augmented Generation (RAG)-enhanced generative AI. The system processes real-time physiological data (ECG, SpO2, blood pressure, temperature, respiratory rate, heart rate variability) alongside electronic medical records and clinical assessments. Developed and validated on 94 patients (74\% aged >65 years) from the Cardio-Thoracic Clinic of University Hospital of Patras, with synthetic data augmentation generated from the MIMIC-IV-ED cohort, the Transformer-based architecture achieves 96.5\% accuracy with Area Under the Curve (AUC) of 0.98 (6h) and 0.96 (24h). DETER outperforms traditional EWS by providing risk stratification with lead times of days to weeks, not hours, and generates continuously updated, personalized deterioration risk scores for proactive intervention and optimized resource allocation. The integration of continuous telemetry with RAG-enhanced generative AI moves monitoring from reactive to predictive and from population-level to personalised. By forecasting physiological trajectories with lead time in days rather than hours, DETER can enable earlier intervention, reduce preventable deterioration, and support both acute and ambulatory care. Prospective validation is warranted to confirm real-world impact and feasibility in diverse healthcare environments.},
  keywords = {Generative AI, clinical deterioration, machine learning, remote patient monitoring},
  issn = {request pending},
  publisher = {Institute of Central Computation and Knowledge}
}