The Use of Several Information Criteria for Logistic Regression Model to Investigate the Effects of Diabetic Drugs on HbA1c Levels

Abstract

HbA1c measurement is an important indicator for checking the diabetes care in a patient. There are several drugs to hold HbA1c in desired level. Diabetic drugs assist the patients for decreasing the glucose level so HbA1c level can be healed. In this article we investigated the effects of diabetic drugs on HbA1c level for type 2 diabetes patients. We implemented variable selection procedures with logistic regression analysis. Particle Swarm Optimization (PSO) was used to minimize the fitness function of the regarding models with several information criteria. According to selection system of the information criteria, we obtained different variable subsets and interpreted the regression models. In application part, we evaluated each selected logistic regression model and identified the common efficient drugs for HbA1c level. Our results demonstrate that Insulin, Metformin, Glyburide, Glipizide and Glimepiride are the joint effective drugs. Also Insulin is the most influential drug to balance the HbA1c level of diabetic patients. © 2018, Scientific Publishers of India. All rights reserved.