Internal Structure of a Disease Name and its Application for ICD Coding

ICD-coding is a complex and difficult task. Coding results vary a great deal depending on each coder's ability. Although the Japanese Standard Disease-Code Master facilitates the coding tasks, it also engenders post-coordination problems derived from combinations of basic diseases (with ICD code) and modifiers. Post-coordination sometimes alters the original ICD code dramatically. To solve this problem, this paper presents a proposal for using internal structures of disease names to correct the ICD code. First, we built an internal structure analyzer, which achieved high (83.7%) accuracy. Results demonstrated that the analyzed output is helpful for precise ICD-coding tasks.