Construction of a semi-automatic ICD-10 coding system.

Area of Science:

• Medical Informatics
• Health Information Management
• Computational Linguistics

Background:

• Manual International Classification of Diseases, 10th Revision (ICD-10) coding is resource-intensive and prone to errors.
• Existing automatic coding approaches have limited practical application.
• There is a need for efficient and accurate automated ICD-10 coding solutions.

Purpose of the Study:

• To develop a practical, automatic ICD-10 coding machine.
• To enhance coding efficiency and quality in daily clinical work.
• To automate the assignment of diagnosis codes using regular expressions.

Main Methods:

• Developed a system using regular expressions (regexps) to map diagnosis descriptions to ICD-10 codes.
• Embedded regexp description models into an upgraded coding system.
• Evaluated system performance using precision (P), recall (R), F-measure (F), and accuracy (A) on two distinct datasets.

Main Results:

• Achieved high precision rates of 89.27% and 88.38% in two testing phases.
• Processed over 160,000 ICD-10 codes in 16 months, reducing coder workload.
• Automatic coding was approximately 100 times faster than manual coding.

Conclusions:

• The developed automatic ICD-10 coding system is effective for practical coding tasks.
• Further research is needed to refine regexp models and explore synthetic approaches for performance enhancement.
• The system demonstrates significant potential for improving healthcare administrative efficiency.