TEXTINFO: a tool for automatic determination of patient clinical profiles using text analysis

Area of Science:

• Medical Informatics
• Natural Language Processing
• Clinical Data Analysis

Background:

• Clinical data is often unstructured in narrative patient documents.
• Extracting meaningful information from clinical narratives is challenging.
• Efficient methods are needed to leverage this data for clinical insights.

Purpose of the Study:

• To develop a method for processing narrative clinical data into structured patient profiles.
• To evaluate the utility of these patient profiles for diagnostic and prognostic applications.
• To demonstrate the application of statistical analysis on extracted clinical features.

Main Methods:

• Utilized grammatical and semantic processing for narrative patient documents.
• Created relational database tables from processed clinical data.
• Matched database retrievals against clinical descriptors to form patient profiles.
• Applied factor analysis and discriminant analysis to patient profiles.
• Analyzed discharge summaries from 57 Digestive Surgery patients.

Main Results:

• Patient profiles were successfully generated from narrative clinical documents.
• The profiles demonstrated utility in defining relationships between diagnoses and clinical findings.
• Profiles were effective for assessing diagnoses by comparing definitions with recorded events.
• Potential for outcome evaluation was shown through classification abilities of clinical signs.

Conclusions:

• Grammatical and semantic processing of clinical narratives enables structured data extraction.
• Generated patient profiles are valuable for diagnosis definition and assessment.
• Clinical sign classification from patient profiles offers potential for outcome evaluation.
• This approach enhances the utility of narrative patient data in healthcare.