Artificial Intelligence in primary care: Intelligent risk predicting platform for non-communicable chronic diseases

Abstract

Chronic Non-Communicable Diseases with multifactorial and non-infectious characteristics in their origin have been a matter of growing concern for society and governments, as they put people at greater risk of complications, disability, and death. They are the leading cause of death worldwide, i.e. a global threat to health. Modern medicine relies heavily on a variety of biochemical, behavioral, clinical, etc. data to advance the chronic diseases field. Machine learning techniques have a wide range of applications, for example for diabetes classifications, phenotypic resistance of whole genomic sequences and heart disease. These approaches, in general, concentrate efforts on a single Chronic Non-Communicable Diseases. Thus, identifying multiple chronic diseases simultaneously is an open challeng. Our study demonstrated that the intelligent risk prediction platform for non-communicable chronic diseases using Multi-Label Classification methods can be used to model multiple Chronic Non-Communicable Diseases. We use ensemble methods developed on top of algorithm adaptation techniques, including Random Forest, Extra Trees and Decision Tree Classifiers. Among the models in the experiment, the Random Forest achieved the best accuracy and F1-score performance with 96.16% and 90.48%, respectively. Our results suggest that the platform is promising and can be used to support medical teams in the early diagnosis of Chronic Non-Communicable Diseases, allowing them to provide adequate treatment options to patients in a timely manner in primary care, as well as contributing to reducing the risk of complications, disability, and deaths by Chronic Non-Communicable Diseases.