JMIR Publications

ABSTRACT

Background:

The accuracy of ICD-10-CM/PCS (International Classification of Diseases, 10th Revision Clinical Modification/Procedure Coding System) coding is crucial for generating correct Taiwan Diagnosis-Related Groups (TwDRGs), as coding errors can lead to financial losses for hospitals.

Objective:

The study aimed to determine the consistency between the artificial intelligence (AI)-assisted coding module and manual coding, as well as identifying clinical specialties suitable for implementing the developed AI-assisted coding module.

Methods:

This study validates the AI-assisted coding module from the perspective of healthcare professionals. The research period commenced in February 2023. The study subjects excluded cases outside of TwDRGs, those with incomplete medical records, and cases with TwDRGs disposals ICD-10-PCS. Data collection was conducted through retrospective medical record review. The AI-assisted module was constructed using a hierarchical attention network (HAN). The verification of the TwDRGs results from the AI-assisted coding model focused on the major diagnostic category (MDC). Statistical computations were conducted using statistical package for the social sciences (SPSS) software, while research variables consisted of categorical variables represented by MDC, and continuous variables represented by the RW of TwDRGs.

Results:

A total of 2,632 discharge records meeting the research criteria were collected from 0February to April 2023. In terms of inferential statistics, Kappa statistics were employed for MDC analysis. The infectious diseases, parasitic diseases and respiratory system had Kappa values exceeding 0.8. Clinical inpatient specialties were statistically analyzed using the Wilcoxon Signed Rank Test. There was no difference in coding results between 23 clinical departments such as Division of Cardiology, Division of Nephrology, and Department of Urology classification personnel.

Conclusions:

For human coders, with the assistance of the ICD-10-CM/PCS AI-assisted coding system, work time is reduced; additionally, strengthening knowledge in clinical documentation improvement (CDI) enables human coders to maximize their role. This positions them to become CDI experts1, preparing them for further career development. Future research will apply the same methodology to validate the ICD10PCS AI-assisted coding module.