Real-time prospective (shadow mode) validation of an AI-based clinical decision support system for predicting 3-month functional outcome in acute stroke: the VALIDATE study protocol
13 affiliations
- 1Stroke Unit, Hospital Universitari Vall d’Hebron, Barcelona, Spain
- 2Stroke Research Group, Vall d’Hebron Institut de Recerca, Barcelona, Spain
- 3Department of Neuroradiology, University of Heidelberg, Heidelberg, Germany
- 4Department of Neurology, Hebrew University Hadassah Medical School, Yerushalayim, Israel
- 5Charité Lab for Artificial Intelligence in Medicine (CLAIM), Charité - Universitätsmedizin Berlin, Berlin, Germany
- 6Digital Health Department, IBM iX GmbH, Berlin, Germany
- 7Simula Metropolitan Center for Digital Engineering AS, Oslo, Norway
- 8Fuenlabrada School of Engineering, Rey Juan Carlos University, Fuenlabrada, Spain
- 9NORA Health, Sant Cugat, Spain
- 10Technological University Dublin, Dublin, Ireland
- 11Digital Health, Empirica Gesellschaft für Kommunikations- und Technologieforschung mbH, Bonn, Germany
- 12ADAPT Research Centre, School of Computer Science and Statistics, Trinity College, Dublin, Ireland
- 13Corporate Member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Department of Neurosurgery, Charité - Universitätsmedizin Berlin, Berlin, Germany
Abstract
Introduction: Reperfusion therapies improve outcomes in acute ischaemic stroke, but hyperacute treatment decisions remain complex and are rarely informed by individualised outcome predictions. Artificial intelligence (AI)-based clinical decision support systems (CDSS) could provide real-time prognostic estimates, yet prospective evidence of their feasibility and performance within routine workflows is scarce. We aim to prospectively evaluate the real-time feasibility, usability and predictive performance of an AI-based CDSS (VALIDATE-CDSS) for individualised outcome prediction in acute stroke care.
Methods and analysis: Prospective, multicentre observational study enrolling consecutive patients with acute ischaemic stroke at three tertiary stroke centres. Clinical management will follow standard practice at the discretion of treating physicians. In parallel, a dedicated researcher will collect patient data in real time and enter them into VALIDATE-CDSS via a mobile application, operating in shadow mode without influencing clinical decisions. The system will generate individualised predictions of 3-month functional outcome (modified Rankin Scale) for four treatment strategies (intravenous thrombolysis, endovascular thrombectomy, combined therapy or no reperfusion) at three sequential time points: baseline clinical data, non-contrast CT and CT angiography. The primary outcome is the real-world feasibility and usability of VALIDATE-CDSS within the hyperacute stroke workflow. Secondary outcomes include predictive performance, agreement between model-suggested and actual treatments, incremental value with increasing data availability and potential bias across predefined subgroups.
Ethics and dissemination: Enrolment began after approval by the ethics committees of all participating centres (Vall d'Hebron Institut de Recerca, PR(AG)432-2023; Ethikkommission der Medizinischen Fakultät Heidelberg, S-687/2023; Helsinki Committee, Hadassah Medical Center-Ein Kerem, HMO-0529-22). Results will be disseminated through peer-reviewed open-access journals and conference presentations. Following open science principles, anonymised data and metadata will be deposited in the Zenodo repository on study completion.
Trial registration number: NCT05622539 .
Related
- An explainable artificial intelligence system for prognostic assessment and decision support in diabetic foot ulcer care: a multicentre derivation and validation study with prospective randomised evaluation
- Development and internal temporal validation of an explainable XGBoost model for predicting clinical pregnancy after frozen embryo transfer
- Longitudinal Clinical and Functional Outcomes in Adult-Onset Delusional Disorder and Schizophrenia: A 10-Year Prospective Comparison Study
- Development and Validation of an Interpretable Machine‐Learning Model for Predicting Suboptimal Clinical Response After Sleeve Gastrectomy: A Multicenter Study
- Predicting acute kidney injury in type B aortic dissection: development and validation of a clinical nomogram.
BibTeX
@article{rubiera2026realtime,
title = {{Real-time prospective (shadow mode) validation of an AI-based clinical decision support system for predicting 3-month functional outcome in acute stroke: the VALIDATE study protocol}},
author = {Marta Rubiera and Martin Bendszus and Ronen R Leker and Adam Hilbert and Ingo Werren and Luis Miguel Lopez-Ramos and Mercedes Ayesta and Que Nguyen and Sussane Bonekamp and Victoria Sala and Hamza Jubran and Claudia Meza and Fatma Shalabi and Yoel Schwartzmann and David Cano and Malte von Tottleben and John Kelleher and Dietmar Frey},
journal = {BMJ Open},
year = {2026},
month = oct,
doi = {10.1136/bmjopen-2026-120511},
url = {https://bmjopen.bmj.com/lookup/doi/10.1136/bmjopen-2026-120511},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Real-time prospective (shadow mode) validation of an AI-based clinical decision support system for predicting 3-month functional outcome in acute stroke: the VALIDATE study protocol
AU - Marta Rubiera
AU - Martin Bendszus
AU - Ronen R Leker
AU - Adam Hilbert
AU - Ingo Werren
AU - Luis Miguel Lopez-Ramos
AU - Mercedes Ayesta
AU - Que Nguyen
AU - Sussane Bonekamp
AU - Victoria Sala
AU - Hamza Jubran
AU - Claudia Meza
AU - Fatma Shalabi
AU - Yoel Schwartzmann
AU - David Cano
AU - Malte von Tottleben
AU - John Kelleher
AU - Dietmar Frey
PY - 2026
DA - 2026/10/07
JO - BMJ Open
DO - 10.1136/bmjopen-2026-120511
UR - https://bmjopen.bmj.com/lookup/doi/10.1136/bmjopen-2026-120511
AB - Introduction: Reperfusion therapies improve outcomes in acute ischaemic stroke, but hyperacute treatment decisions remain complex and are rarely informed by individualised outcome predictions. Artificial intelligence (AI)-based clinical decision support systems (CDSS) could provide real-time prognostic estimates, yet prospective evidence of their feasibility and performance within routine workflows is scarce. We aim to prospectively evaluate the real-time feasibility, usability and predictive performance of an AI-based CDSS (VALIDATE-CDSS) for individualised outcome prediction in acute stroke care. Methods and analysis: Prospective, multicentre observational study enrolling consecutive patients with acute ischaemic stroke at three tertiary stroke centres. Clinical management will follow standard practice at the discretion of treating physicians. In parallel, a dedicated researcher will collect patient data in real time and enter them into VALIDATE-CDSS via a mobile application, operating in shadow mode without influencing clinical decisions. The system will generate individualised predictions of 3-month functional outcome (modified Rankin Scale) for four treatment strategies (intravenous thrombolysis, endovascular thrombectomy, combined therapy or no reperfusion) at three sequential time points: baseline clinical data, non-contrast CT and CT angiography. The primary outcome is the real-world feasibility and usability of VALIDATE-CDSS within the hyperacute stroke workflow. Secondary outcomes include predictive performance, agreement between model-suggested and actual treatments, incremental value with increasing data availability and potential bias across predefined subgroups. Ethics and dissemination: Enrolment began after approval by the ethics committees of all participating centres (Vall d'Hebron Institut de Recerca, PR(AG)432-2023; Ethikkommission der Medizinischen Fakultät Heidelberg, S-687/2023; Helsinki Committee, Hadassah Medical Center-Ein Kerem, HMO-0529-22). Results will be disseminated through peer-reviewed open-access journals and conference presentations. Following open science principles, anonymised data and metadata will be deposited in the Zenodo repository on study completion. Trial registration number: NCT05622539 .
SN - 2044-6055
SN - 2044-6055
ER -