Similarity Matching for Workflows in Medical Domain Using Topic Modeling

Abstract

The healthcare industry is a complex domain involving a range of different interests including individual patients, medical service providers, hospitals, clinics, and support organizations, including insurance, testing, and research organizations. Given increasing patient loading, accelerating expansion of domain knowledge, advent of online healthcare and the greater number of institutions now participating in medical decision making, the challenges of its management are daunting. Handling of data has already evolved from individual care providers operating in isolation to varied approaches more reliant on automated systems. Expert systems have become more welcome, but in isolation, provide limited assistance. We develop a corpus of automatically captured information using workflow technology of patient history, testing and treatment along with disease research, symptoms, and treatment. We present an automated method using topic modeling and knowledge-based similarity measurements to suggest meaningful similarities between patients and applicable diagnoses.