[Automatic text analysis of medical records]

Area of Science:

• Medical Informatics
• Natural Language Processing
• Clinical Data Analysis

Context:

• Computerized medical records (PROMED) contain vast amounts of clinical narrative data.
• Automated analysis of this data can yield valuable medical insights.
• Existing methods may lack the ability to extract nuanced clinical information without pre-programmed medical knowledge.

Purpose:

• To evaluate the capability of a novel computer program (LOGSTORY) to automatically analyze clinical narratives.
• To assess LOGSTORY's ability to extract medical knowledge, including symptoms, signs, and etiology, without prior medical knowledge.
• To quantify similarities and differences between various clinical states.

Summary:

• LOGSTORY analyzed 5,041 patients and 14,323 diagnoses from PROMED, focusing on general practice and occupational medicine cases.
• The program successfully reproduced and quantified key aspects of diabetes mellitus, obesity, and lung diseases.
• It demonstrated the ability to recognize and quantify similarities and differences between clinical states.

Impact:

• PROMED-LOGSTORY can automatically extract medical knowledge from clinical narratives.
• The system may serve as a valuable tool for medical self-evaluation, peer review, and quality control.
• This approach offers potential for advancing clinical research and medical education.