Real-time clinical analytics at scale: a platform built on large language models-powered knowledge graphs

Rui Li, Shuang Cao, Louise Ruihua Liu, Alexandre Duprey

Abstract:

Objectives
The increasing volume of clinical trial documents presents a significant challenge for biomedical researchers who must analyze vast amounts of unstructured and structured data. Traditional methods are no longer feasible given the scale and complexity of modern clinical trials.

Materials and Methods
We introduce a large-scale clinical analytics platform, ClinicalMind, that integrates Large Language Models (LLMs) with Knowledge Graph technology to perform real-time clinical analytics over 110 000 clinical documents and 60 000 electronic medical records. The system employs a 2-phase graph update strategy and hardware acceleration to increase the accuracy and speed.

Results
Our platform achieves an average query delay of 1.7 seconds with high accuracy (BLEU score: 0.85, ROUGE score: 0.92). The system can process and analyze thousands of clinical documents in real-time, significantly outperforming existing methods.

Discussion
These results demonstrate that combining LLMs with a continuously updated knowledge graph enables scalable, low-latency clinical analytics across large and heterogeneous data sources. The observed performance gains highlight the potential of this approach to support real-time clinical decision-making and large-scale evidence synthesis, addressing key limitations of existing document-centric and retrieval-based methods.

Conclusion
We demonstrate that our platform offers an efficient, scalable solution for real-time clinical analytics, enabling rapid analysis of large-scale clinical document collections.