A high throughput semantic concept frequency based approach for patient identification: a case study using type 2...

Area of Science:

• Biomedical Informatics
• Clinical Data Science

Background:

• Automated patient identification for specific phenotypes is an emerging research area.
• A generalizable, automatic approach is urgently needed for high-throughput patient identification.

Purpose of the Study:

• To develop a novel method for automatic patient identification using electronic clinical notes.
• To evaluate the utility of integrating SNOMED semantic knowledge into patient identification.
• To investigate the effectiveness of combining Natural Language Processing (NLP), machine learning, and ontology.

Main Methods:

• Utilized Mayo Clinic electronic clinical notes.
• Applied Support Vector Machine (SVM) algorithm on SNOMED concept units extracted from T2DM case/control notes.
• Calculated precision, recall, and F-score for performance evaluation.

Main Results:

• Achieved an F-score exceeding 0.950 for both case and control groups.
• Identified 'Disease or Syndrome' semantic type concept units as most informative for patient identification.
• Demonstrated that coarse-level concepts provide sufficient information for classifying T2DM cases and controls.

Conclusions:

• The proposed method effectively identifies patients with specific phenotypes.
• Leveraging SNOMED semantic knowledge enhances patient identification accuracy.
• NLP, machine learning, and ontology integration offers a robust approach for clinical informatics.