Clinical Report: AI-Driven Early Detection of Infectious Diseases in Dutch Primary Care

Overview

This study presents ERNIE, an AI-based natural language processing framework that autonomously detects emerging infectious disease clusters from unstructured primary care text data. Validated on Dutch datasets, ERNIE identified early COVID-19-like clusters prior to confirmed cases and detected RSV and simulated West Nile Virus patterns with high recall and precision.

Background

Emerging infectious diseases require timely detection to enable effective public health responses. Traditional surveillance systems rely heavily on structured data such as diagnostic codes and often miss early, non-specific clinical signals. Dutch general practitioners document rich clinical narratives in free-text fields, offering a valuable but underutilized data source for early outbreak detection. Advanced NLP models like BERT enable extraction of meaningful insights from such unstructured data, potentially overcoming limitations of code-based surveillance.

Data Highlights

Key Findings

• ERNIE detected early COVID-19-like symptom clusters before the first confirmed case in the Netherlands.
• It identified respiratory syncytial virus (RSV) indicative patterns during the 2021 surge.
• Simulated West Nile Virus cases were successfully flagged by the framework.
• ERNIE maintained stability during control periods without outbreaks, demonstrating robustness.
• The framework operates without reliance on diagnostic codes or laboratory confirmations, using only routine clinical text.
• It integrates autoencoders, clustering, key term extraction, and visualization for interpretable surveillance.

Clinical Implications

ERNIE offers a scalable, autonomous tool for early detection of emerging infectious diseases directly from primary care narratives, potentially enabling faster public health interventions. Its disease-agnostic approach allows identification of atypical clinical patterns before formal diagnoses or coding exist, addressing a critical gap in current surveillance systems. Incorporating such AI-driven methods into routine primary care monitoring could reduce clinical and economic burdens associated with delayed outbreak recognition.

Conclusion

The ERNIE framework demonstrates that AI-driven analysis of unstructured primary care text can effectively detect emerging infectious disease clusters early and autonomously. This approach holds promise for enhancing infectious disease surveillance and supporting timely healthcare responses.

References

1. Study Source 2024 -- Utilizing AI for Early Identification of Infectious Diseases in Dutch Primary Care with BERT and ERNIE