Real-world data in rheumatoid arthritis: patient similarity networks as a tool for clinical evaluation of disease activity

Abstract

Abstract Hospital databases provide complex data on individual patients, which can be analysed to discover patterns and relationships. This can provide insight into medicine that cannot be gained through focused studies using traditional statistical methods. A multivariate analysis of real-world medical data faces multiple difficulties, though. In this work, we present a methodology for medical data analysis. This methodology includes data preprocessing, feature analysis, patient similarity network construction and community detection. In the theoretical sections, we summarise publications and concepts related to the problem of medical data, our methodology, and rheumatoid arthritis (RA), including the concepts of disease activity and activity measures. The methodology is demonstrated on a dataset of RA patients in the experimental section. We describe the analysis process, hindrances encountered, and final results. Lastly, the potential of this methodology for future medicine is discussed.

Keywords

• Patient similarity network
• Local representativeness
• LRNet
• Rheumatoid arthritis
• Medical data