Real-time peri-operative data integration and clinical decision support: an open interoperable platform for patient data.

Abstract

Introduction: Modern intensive care units (ICU) and operating rooms (OR) generate vast quantities of physiological data, yet this information is often underused due to architectural barriers and lack of interoperability, a failure to integrate data from multiple devices. While imaging data has achieved global standardisation, continuous physiological time series remain fragmented and inaccessible for scientific use, limiting multicentric collaboration and the development of robust artificial intelligence (AI) models.

Methods: We developed an open, interoperable peri-operative data platform compliant with General Data Protection Regulation. The platform integrates data from multiple sources, including patient monitors, syringe pumps, ventilators, EEG systems, and laboratory values, using Apache Kafka for real-time data streaming and a hybrid long-term storage architecture. Data governance was ensured through on-premise storage, anonymisation protocols, and ethical approvals.

Results: Since March 2019, the platform has recorded data from 118 915 patients across 77 OR and ICU beds, encompassing waveform and numeric channels. The database includes data from various clinical specialties, such as bowel surgery, otolaryngology, neurosurgery, gynaecology, interventional neuroradiology, and orthopaedics. It encompasses cardiac output monitoring (3458 patients), ventilation data (56 212 patients), advanced ECG analysis (78 091 patients), and EEG recordings (9862 patients). Syringe pump data, collected since February 2023, includes 7096 cases.

Discussion: The platform demonstrates the feasibility of creating a scalable, vendor-neutral infrastructure for peri-operative data, enabling both real-time clinical applications and long-term research. By adhering to open standards, it supports secure, cross-institutional collaboration while preserving data integrity. However, challenges remain in standardising data quality and expanding data volume to improve model generalisability.

Conclusions: This work provides a reproducible blueprint for building interoperable peri-operative data platforms, enabling both real-time clinical decision support and large-scale research. By combining real-time acquisition with durable storage and open standards, the platform transforms fragmented physiological data into a shared resource for precision medicine and clinical equity.