Leveraging Non-Clinical Factors and Machine Learning for Improved Tuberculosis Prediction in Resource-Limited Areas

Abstract

Despite being curable and treatable in the majority of modern nations, tuberculosis (TB) remains a major public health concern and a top cause of infection-related deaths globally. Non-clinical factors like biological, socio-economic, and environmental factors are relevant to prediction and prevention, as relying only on clinical signs does not encompass other risk factors that impact tuberculosis transmission. This study's main goal is to examine the application of machine learning (ML) techniques while emphasizing the impact of non-clinical elements. Data for the study was collected from a variety of public and private healthcare facilities in a few chosen states in the Niger Delta region of Nigeria. The ability of three machine-learning classifiers is evaluated to predict and detect dengue fever for the application in TB. RIPPER was found to be a balanced classifier due to its high F-Measure and ROC-AUC scores. This study focuses on non-clinical factors that affect the spread of TB in addition to the significance of ML with its inclusion of non-clinical factors with its approach to prediction.