Clinical Application of Detecting COVID-19 Risks: A Natural Language Processing Approach

Area of Science:

• Computational linguistics
• Public health informatics
• Machine learning for healthcare

Background:

• Detecting COVID-19 related factors is crucial but challenging.
• Existing named entity recognition models have limitations in scope.
• Non-clinical factors like social determinants of health (SDoH) are vital for infectious disease research.

Purpose of the Study:

• To develop a generalizable machine learning approach for recognizing a broad range of clinical risk factors and SDoH.
• To improve upon existing methods for identifying health-related entities in text data.
• To provide a tool for researchers and clinicians studying infectious diseases.

Main Methods:

• Utilized a combination of deep neural networks, including BiLSTM-CNN-CRF.
• Incorporated a transformer-based embedding layer for enhanced feature representation.
• Trained and evaluated the model on a dataset derived from PubMed articles concerning COVID-19.

Main Results:

• The proposed approach demonstrated superior performance compared to existing methods.
• Achieved performance gains of approximately 1-5% in macro- and micro-average F1 scores.
• Successfully recognized a larger number of clinical risk factors and SDoH relevant to COVID-19.

Conclusions:

• The novel machine learning method offers improved accuracy in identifying clinical risks and SDoH.
• This approach can assist clinical practitioners and researchers in gaining precise information.
• The developed pipeline serves as a valuable tool for pandemic response and future preparedness.