Disease coding is the process of transforming diagnosis names to standard ICD(International Classification of Diseases) codes. Due to the huge amount of workload and the low efficiency of manual coding, it's necessary to achieve the automation of disease coding. This paper presents a method for automated coding of disease. Specifically, we present ICD as documents with a text modeling approach, and then get the most relevant ICD code of a diagnosis name with the text correlation measure. The experiments proves that our method has a good accuracy, it's very efficient and easy to switched among classification levels, could be widely applied to various types of data analysis scenarios.