Analysis of ICD-10 Coding Errors in 43 Files Database Systems for Medical Record Department Using Data Mining Techniques

Abstract

This research aims to analyze coding errors in 43 files database systems for medical records department using data mining techniques. We compare performance of decision tree (C4.5 algorithm) and Naïve Bayes algorithms in analyzing and classifying the errors. In addition, association rules for analyzing ICD-10 coding errors among the data was also generated by the Apriori algorithm. 33,862 records of data used in analysis were collected during October 1^st, 2016 to September 30^th, 2017. Weka 3.8.1 was used to analyze and generate models. The results, when evaluating the classification efficiency with 10-folds cross validation method, showed that the classification accuracy of the decision tree was 90.16%, and that of the Naïve Bayes algorithm was 89.87%. When analyzing the data using Apriori algorithm, the result showed that the error code found the most was error code B4 which is referred to the outpatient code of the vaccine (Z23.0 -Z27.9) with no code of physical and health examination Z00.0-Z00.9, Z01.0-Z01.9, Z02.0-Z02.9.