Improving Quality of ICD-10 (International Statistical Classification of Diseases, Tenth Revision) Coding Using AI:...

Area of Science:

• Health Informatics
• Medical Coding Systems
• Artificial Intelligence in Healthcare

Background:

• Clinical coding, essential for health systems and data integrity, is complex and error-prone.
• Modern systems like ICD-11 increase coding challenges, limiting current understanding of Computer-Assisted Coding (CAC) tool effectiveness.
• Existing user studies are limited, necessitating further research into CAC's role in reducing coding burden and enhancing quality.

Purpose of the Study:

• To quantitatively and qualitatively assess the usefulness of the Easy-ICD CAC system for recommending ICD-10 codes.
• To determine if the Easy-ICD tool can decrease the workload for clinical coders.
• To evaluate the potential of the CAC system to improve the accuracy of clinical code assignment.

Main Methods:

• A crossover randomized controlled trial design was employed to compare coder performance with and without the CAC tool.
• Coder performance was measured by the time taken to assign codes to clinical texts.
• Coding quality, defined as the accuracy of code assignment, was also a key performance metric.

Main Results:

• The study is expected to quantify the effectiveness of the CAC system against manual coding in terms of both time efficiency and accuracy.
• Positive results would indicate CAC tools can alleviate healthcare staff burden, supporting AI integration in coding practices.
• Expected results are anticipated for publication in summer 2024.

Conclusions:

• The user study aims to enhance understanding of CAC systems' real-world impact on clinical coding efficiency and quality.
• Findings may offer insights into leveraging clinical text mining for reducing coder burden.
• The research seeks to lower adoption barriers for modern coding systems like ICD-11.