Building an automated SOAP classifier for emergency department reports

Area of Science:

• Medical Informatics
• Natural Language Processing
• Clinical Documentation

Background:

• Clinical notes require structured data for improved information extraction.
• The Subjective, Objective, Assessment, Plan (SOAP) framework is a standard for clinical progress notes.
• The utility of the SOAP framework for automated information extraction remains underexplored.

Purpose of the Study:

• To evaluate the applicability of the SOAP framework to clinical narratives.
• To develop and assess automated SOAP classifiers for sentence classification in emergency department reports.

Main Methods:

• Manual application of the SOAP framework to emergency department reports.
• Training and evaluation of SOAP classifiers using diverse linguistic features.
• Quantitative analysis of classification performance using Cohen's kappa and F(1) scores.

Main Results:

• High inter-rater agreement (Cohen's kappa > 0.70) for manual SOAP classification.
• Automated classifiers achieved high performance: F(1) scores of 93.9% (Subjective), 94.5% (Objective), 75.7% (Assessment), and 77.0% (Plan).

Conclusions:

• The SOAP framework is a viable structure for clinical narratives.
• Automated SOAP classification shows significant potential for enhancing clinical information extraction.
• Further expansion and application of SOAP classification in information extraction tasks are warranted.