An open natural language processing (NLP) framework for EHR-based clinical research: a case demonstration using the...

Area of Science:

• Clinical Natural Language Processing (NLP)
• Translational Research
• COVID-19 Research

Background:

• Clinical NLP model adoption is limited by process heterogeneity and human factors.
• Developing robust NLP models requires multi-site data for generalizability.
• Translational research faces challenges in standardizing data and processes across institutions.

Purpose of the Study:

• To report on developing an NLP solution for Coronavirus Disease 2019 (COVID-19) signs and symptom extraction.
• To highlight the benefits of using multi-site data in NLP model development.
• To emphasize the need for federated annotation and evaluation in multi-site NLP projects.

Main Methods:

• Developed an NLP solution within an open NLP framework.
• Utilized a subset of data from participating sites in the National COVID Cohort (N3C).
• Employed both symbolic and statistical methods for NLP model development.

Main Results:

• Demonstrated the empirical benefits of multi-site data for both symbolic and statistical NLP methods.
• Identified several pitfalls encountered during the development process in a multi-site setting.
• Showcased the utility of an open NLP framework for collaborative research.

Conclusions:

• Multi-site data significantly benefits the robustness and generalizability of clinical NLP models.
• Federated annotation and evaluation strategies are essential to overcome challenges in multi-site NLP development.
• The findings support the use of open frameworks and collaborative approaches for advancing clinical NLP in translational research.