How to implement a decision support for digital health: Insights from design science perspective for action research in tuberculosis detection

Abstract

Association rule mining has gained much popularity in facilitating disease diagnosis and the healthcare industry's decision-making process. The cases of Drug Resistance Tuberculosis (DR-TB) have substantially increased in recent times and the factors driving its rapid rise have not been clearly identified. This research explores the Frequent Pattern (FP) Growth algorithm to identify recurring relationships, disease co-occurrences, and generate some very interesting diagnostics rules for DR-TB. This is achieved through data extraction from patients' TB health database, transforming raw data into a knowledge discovery system that provides a good idea for further exploration through medical research to clarify the unknown patterns from the obtained result. The FP growth algorithm efficiently generates the frequent symptoms associated with pulmonary tuberculosis and drug-resistance tuberculosis by generating the association rules at minimum confidence of 80%. The analysis performed is knowledgeable and conforms to the clinical practices and design science research processes.