Using Routine Data to Categorize Poor Control Diabetic Patients: An Application of Cluster Analysis Technique

Background: Diabetes mellitus is one of the most important causes of morbidity and mortality all over the world. Chronic high blood glucose is the root cause of developing future micro and macro vascular complications. In current study, we analyzed poorly controlled patients’ information by clustering method in order to find more homogenous sub-groups of patients for planning for further interventions. Methods: This study was conducted based on medical records of diabetic patients admitted to 23 health care centers in four cities of Iran from 2010 to 2014. Demographic data and physical and biochemical measurements were extracted and analyzed using two-steps cluster analysis method. Results: Three distinct clusters were derived from 2087 eligible cases. Cluster 1 totally consisted of men without any apparent risk factors. Members of clusters 2 and 3 were illiterate women with obesity. Dyslipidemia was a prominent characteristic in members of cluster 2. Conclusion: Masculinity would play the main role in diabetes control. Meanwhile, aggregation of obesity, illiteracy and/or dyslipidemia in diabetic women predispose them to poor control condition. Based on these findings, we will be able to plan interventions that are more appropriate for identified groups.