CLINICAL DIAGNOSES FOLLOWING ZIPF'S LAW

Abstract

We examine the statistical properties of data relating to clinical diagnoses extracted from the medical database of our hospital. Specifically, we analyze all diagnoses given to all patients (in- and out-patients). The data consisted of both data input to to computers as Japanese names, and the data input ICD10 (International Classification of Diseases 10th Revision). We adapted the Zipf approach to analyzing the frequencies of clinical diagnoses for these two data groups. We found that both group types have the inverse-power relationship between the rank order of diagnoses and the frequency of the appearance of these diagnoses. (This relationship is called Zipf's law, which is observed in natural language.) Though the reason why these sets follow Zipf's law is unknown, we speculate that the complex interaction between doctor and patient is the cause for adherence to Zipf's law.