A Bayesian Machine Learning Model for Predicting Uncontrolled Type 2 Diabetes Mellitus

Type 2 diabetes mellitus (T2DM) is a chronic disease with multiple causes that represents a health and social problem. In the early stages, it may not be symptomatic, and late diagnosis and inadequate management can lead to severe complications, such as vision loss, heart attack, kidney failure, and limb amputation. Those complications, that impact the quality of life and cause premature death in patients, are highly related to their T2DM being uncontrolled. Although there is evidence on the physiopathology of metabolic imbalance, little is known about the many potential external variables, such as socioeconomic status, lifestyle, health knowledge and beliefs and environmental influences that are related to hyperglycemia. Therefore, this study aims to use a simple machine learning model applied to a dataset generated from a highly multi-factorial, multidisciplinary questionnaire applied to a group of diabetics to predict the profiles of those patients most likely, or at risk, of having uncontrolled T2DM. As a result of our research, we propose an exploratory framework which could help in evaluating and predicting complications related to inadequate management of T2DM. These insights indicate that, by providing clear information and adequate education, patients can be helped to understand the importance of adopting a holistic approach in managing diabetes, incorporating both pharmacological treatment and lifestyle changes. This will enable them to have greater control over their health and improve their long-term quality of life.