An automatic system to identify heart disease risk factors in clinical texts over time

Area of Science:

• Clinical Informatics
• Natural Language Processing (NLP)
• Cardiovascular Disease Research

Background:

• Heart disease is a leading cause of death, necessitating effective risk factor identification for prediction and prevention.
• Existing studies have not comprehensively identified all potential heart disease risk factors.
• The 2014 i2b2 challenge focused on identifying heart disease risk factors in longitudinal clinical texts.

Purpose of the Study:

• To develop a system for identifying heart disease risk factors and tracking their progression in clinical texts.
• To address the need for comprehensive risk factor identification in patient medical histories.
• To participate in and achieve high performance in the 2014 i2b2 clinical NLP challenge (track 2).

Main Methods:

• Developed a hybrid pipeline system combining machine learning and rule-based approaches.
• Utilized natural language processing (NLP) techniques to extract risk factors, medications, and disease attributes.
• Applied the system to a challenge corpus of longitudinal patient medical records.

Main Results:

• The hybrid system achieved an F1-score of 92.68% on the challenge corpus.
• The system ranked first among participants without additional annotations in the 2014 i2b2 challenge.
• Demonstrated effective identification of medically relevant information for heart disease risk.

Conclusions:

• The developed hybrid NLP system is highly effective for identifying heart disease risk factors in clinical text.
• This approach can aid in the preliminary steps of heart disease prediction and prevention.
• The system's performance highlights the potential of NLP in extracting critical clinical information over time.