Identification of Multimorbidity Patterns in Rheumatoid Arthritis Through Machine Learning

Area of Science:

• Rheumatology and computational epidemiology.

Background:

• Understanding the complex interplay of chronic conditions in rheumatoid arthritis (RA) is crucial for effective patient care.
• Multimorbidity patterns in RA patients remain poorly defined, necessitating advanced analytical approaches.

Purpose of the Study:

• To identify and define the prevalence of multimorbidity patterns in rheumatoid arthritis (RA) using machine learning techniques.
• To investigate the association between RA and specific clusters of co-occurring chronic conditions.

Main Methods:

• Utilized machine learning, specifically exploratory factor analysis, on large insurance (MarketScan) and healthcare (VHA) databases.
• Constructed RA and matched non-RA cohorts, analyzing 44 chronic conditions identified via diagnosis codes.
• Employed conditional logistic regression to assess the relationship between RA and identified multimorbidity patterns.

Main Results:

• Identified predominant multimorbidity patterns including cardiopulmonary, cardiometabolic, and mental health/chronic pain disorders.
• These patterns were consistent across RA and non-RA cohorts, sexes, and databases.
• RA patients exhibited significantly higher odds for all identified multimorbidity patterns, with mental health and chronic pain disorders showing the strongest association.

Conclusions:

• Cardiopulmonary, cardiometabolic, and mental health/chronic pain disorders are prevalent multimorbidity patterns overrepresented in RA.
• Recognizing these distinct patterns is a critical step towards a comprehensive, holistic management strategy for RA.
• Further research is warranted to explore the clinical and predictive value of these identified multimorbidity patterns in RA care.