Journals Proceedings

Abstract

The International Statistical Classification of Diseases and Related Health Problem 10th Revision (ICD-10) was used extensively in all public hospitals in Thailand. All outpatients and inpatient diagnosis must be coded to ICD-10. Every year 250 million ICD-10 codes were sent to the information department at Ministry of Public Health Thailand. Manual coding by clinical coders could not be completed in some hospitals. Semi-automated ICD-10 coding software is needed to overcome this obstacle. Development of diagnosis directory and algorithm to facilitate physician diagnosis mapping with concepts in ICD-10 is the first step before semi-automated ICD-10 coding software development. The diagnosis directory in this work was built upon two sources of diagnosis list. The first source is the previous diagnosis list from the author research work in 2009. The second source is the diagnosis list built from ICD-10 alphabetical index. Final check of the diagnosis directory was performed by comparison with the Unified Medical Language System. The mapping algorithm steps are; 1) abbreviation conversion 2) stop words removed or replaced 3) spell checking 4) word counting and selection of appropriate set of diagnosis statement 5) similarity measurement 6) output of results. The diagnosis directory contains 43,331 diagnosis statements and could be divided into 10 subsets. The algorithm was tested using OPD diagnosis data from one community hospital of the Ministry of Public Health. The data contain 400 diagnosis statements. The algorithm could map 252 (63.0%) original diagnosis statements to diagnosis directory with 152 (38.0%) exact match (Cosine similarity = 1.0). this finding mean that in the future we can use the algorithm to reduce 63% of the manual work done by clinical coders.