Wright Adam, Bates David W
Brigham and Women's Hospital, Boston, MA, USA.
Appl Clin Inform. 2010;1(1):32-37. doi: 10.4338/ACI-2009-12-RA-0023.
Many natural phenomena demonstrate power-law distributions, where very common items predominate. Problems, medications and lab results represent some of the most important data elements in medicine, but their overall distribution has not been reported. OBJECTIVE: Our objective is to determine whether problems, medications and lab results demonstrate a power law distribution. METHODS: Retrospective review of electronic medical record data for 100,000 randomly selected patients seen at least twice in 2006 and 2007 at the Brigham and Women's Hospital in Boston and its affiliated medical practices. RESULTS: All three data types exhibited a power law distribution. The 12.5% most frequently used problems account for 80% of all patient problems, the top 11.8% of medications account for 80% of all medication orders and the top 4.5% of lab result types account for all lab results. CONCLUSION: These three data elements exhibited power law distributions with a small number of common items representing a substantial proportion of all orders and observations, which has implications for electronic health record design.
许多自然现象呈现幂律分布，其中极为常见的项目占主导地位。问题、药物和实验室检查结果是医学中一些最重要的数据元素，但其总体分布情况尚未见报道。目的：我们的目的是确定问题、药物和实验室检查结果是否呈现幂律分布。方法：回顾性分析2006年和2007年在波士顿布莱根妇女医院及其附属医疗机构至少就诊两次的100,000名随机选取患者的电子病历数据。结果：所有这三种数据类型均呈现幂律分布。最常出现的12.5%的问题占所有患者问题的80%，最常用的11.8%的药物占所有药物医嘱的80%，最常见的4.5%的实验室检查结果类型占所有实验室检查结果。结论：这三个数据元素呈现幂律分布，少数常见项目占所有医嘱和检查结果的很大比例，这对电子健康记录设计具有启示意义。
