HYBRID STREAMING AND BATCH INTELLIGENCE FRAMEWORK FOR REAL-TIME HEALTHCARE ANALYTICS ON DISTRIBUTED CLOUD DATA PLATFORMS

Abstract

Hybrid streaming and batch intelligence architecture is becoming the backbone of the clinical data platform as it strives to combine low-latency event processing with retrospective learning from very large volumes of longitudinal clinical data. The review examines peer-reviewed journal publications from 2015 to the present on cloud, fog, Internet of Things (IoT), and critical-care analytics architectures and their application to distribute healthcare intelligence. The reviewed evidence indicates that streaming components can improve responsiveness in physiological monitoring, seizure detection, prediction of acute kidney injury, and monitoring of deterioration in intensive care. For cohort construction, model training, calibration, population evaluation and auditability, batch components are still needed. The literature, however, has been split into infrastructure studies, clinical prediction studies, and database-based benchmarking. The major gaps found are limited treatment of concept drift, lack of reporting of end-to-end latency, limited institutional validation and incomplete models of governance for the continuous updating of clinical intelligence. Architectures are therefore required that integrate paradigms of streaming inference, batch retraining, privacy-preserving federation, and interpretable decision support, all in distributed cloud data platforms, with operational monitoring.