Statistical and Machine Learning Approaches for Clinical Decision on Drug Usage in Diabetes with Reference to Competence and Safeness

Abstract

Diabetes is a chronic disease that requires patient-centered treatment. The physician strategy for treatment of diabetes varies from one patient to another. Using the clinical parameters and the evidence of diabetes at various group of people are to be treated with the drugs that provide significant changes over period of time. In this work, safety and efficiency of drug that is used for diabetes and to provide justification using statistical approach is proposed. The benefits and harm of various drugs are represented as null hypothesis and alternate hypothesis using two-tailed t test (unpaired hypothesis testing). The drugs specified are given periodically at various weeks so that the effect of each drug is identified with clinical parameters and it is summarized. The various medications that are to be imposed on various groups of people and respected hypothesis values are calculated. The post hoc power, evaluation of p value that specify the significant change in the clinical parameters are observed. With the help of this p value and the hypothesis testing, it recommends the correct specification of drugs. The drug combination such as sulfonyl urea (glibenclamide 5 mg), sulfonyl urea + sitagliptin, sulfonyl urea + vildagliptin, metformin, metformin + sitagliptin, metformin + vildagliptin were used in this study. The above drugs are given to various groups to find out the effectiveness of drug usage in diabetes. The idea is implemented with both manual and automated approach of handling patient report and to find their significant approach and thereby to provide conclusion of the drug usage for diabetes.