An Algorithmic Approach to Create Bi-directional Mapping Files Between ICD-10 and ICD-10-AM

Abstract

Given that the health industry uses varying clinical classification systems across countries, it is important to have mappings between these classification systems to enable comparison and statistical analysis of healthcare data across the borders. This paper discusses an algorithmic technique that facilitates the creation of mapping files between an international disease classification and a country-specific extension of the said international classification. The algorithm is tested by creating maps between ICD-10 (International Statistical Classification of Diseases and Related Health Problems 10th Revision) and its Australian Modification (ICD-10-AM). This algorithmic approach leverages Elasticsearch which is a full-text search engine that enables finding the closest lexical match between sentences. The result for ICD-10 to ICD-10-AM is 99.96% sensitivity, 100% specificity with an f-score value of 99.98% while ICD-10-AM to ICD-10 mapping has 99.58% sensitivity, 64.44% specificity and f-score value of 99.75%.