Infection risk in rheumatoid arthritis patients treated with biologic and targeted synthetic DMARDs: prediction models and use of machine learning.

Infection risk is an important consideration during treatment decision-making in rheumatoid arthritis (RA), including at the initiation of biologics or targeted synthetic disease-modifying anti-rheumatic drugs. A number of risk prediction models for serious infection associated with RA treatment currently exist, derived from different patient populations and using conventional statistical approaches. This review describes and compares current prediction models, evaluates their limitations, and explores the potential of machine learning (ML) to enhance current risk prediction. Important aspects related to ML application, including data accuracy, validation and generalizability, interpretability, and legal and regulatory approval, are discussed.