Measuring diversity in medical reports based on categorized attributes and international classification systems

Area of Science:

• Medical Informatics
• Health Data Science
• Clinical Terminology

Background:

• Narrative medical reports lack standardized terminology, hindering statistical analysis and decision-making.
• Current medical reports often contain insufficient data for robust processing.
• The need for standardized terminology in medical reporting is critical for improving data utility.

Purpose of the Study:

• To propose a novel method for measuring diversity in medical reports across languages.
• To compare the diversity of terminology between narrative and structured medical reports.
• To map medical report attributes and terms to established classification systems like SNOMED CT and ICD-10.

Main Methods:

• Developed a new method based on f-diversity to quantify terminology diversity in medical reports.
• Utilized SNOMED CT and ICD-10 for mapping attributes and terms to standardized codes.
• Employed Gini-Simpson and Number of Categories f-diversities to compare report types using attributes from the Minimal Data Model for Cardiology (MDMC).

Main Results:

• Compared 110 Czech narrative and 1119 structured medical reports, finding significant differences in attribute categories and terms.
• Over 60% of MDMC attributes were successfully mapped to SNOMED CT.
• Narrative reports exhibited greater terminology diversity than structured reports for most MDMC attributes, except 'Allergy'.

Conclusions:

• Non-standardized terminology in narrative medical reports impedes mapping to classification systems and complicates computer processing.
• Higher diversity in narrative reports can lead to information loss during data processing.
• Implementing standardized terminology is essential for complete, accessible patient information and improved healthcare delivery.