Analysis of diabetic medical information using the datamart of electronic medical record system

本研究将糖尿病患者的临床资料，如病史、体格检查及实验室检查等，整合成一个以医院为基础的数据集市，以分析糖尿病的临床病程。将该模板应用于电子病历系统中，可以有效地将医疗信息集成到数据集市中。数据集市中的总记录数超过1000万条，使用结构化查询语言（SQL），我们可以轻松地提取必要的数据。我们展示了题为“2型糖尿病新患者血糖控制预测因素的识别”，“住院治疗后血糖控制相关因素”，我们还利用数据挖掘方法分析了预测住院6个月后血糖控制的因素（关联规则、决策树方法、神经网络等）。我们发现，数据挖掘技术是更强大的方法来分析复杂的疾病，如糖尿病比传统的多变量分析。

We have integrated the clinical information such as clinical history, physical examinations and laboratory data into a hospital-based datamart for the analysis of clinical course of diabetes mellitus. Using the template in the electronic medical recored system, we could effectively integrated medical information into the datamart. The number of total records put in the datamart was more than 10 million, and using the structured query language (SQL), we could easily extract the necessary data.We presented the data entitled "Identification of factors predicting glycemic control in new patients with type 2 diabetes", "Factors associated with glycemic control after an inpatient program", "Seasonal fluctuation of glycosylated hemoglobin in Japanese diabetic patients" et al. in the international meeting.We also analyzed the factors which predict glycemic control 6 months after the admission to the inpatient clinic using data-mining methods (association rules, decision-tree methods, neural network etc.). We found that data-mining techniques are more powerful methods to analyze complicated disease like diabetes mellitus than conventional multivariate analysis.