Using Unsupervised Machine Learning Methods to Cluster Comorbidities in a Population‐Based Cohort of Patients With Rheumatoid Arthritis
Authors' conclusion
Despite the higher comorbidity burden in patients with RA compared to the general population, clustering comorbidities did not identify substantial differences in comorbidity patterns between the RA and non-RA cohorts.
Study context
- Population
- Residents of 8 Minnesota counties with prevalent rheumatoid arthritis (RA) as of January 1, 2015, and matched non-RA comparators (72% female; mean age 63.1 years)72% female
- Participants
- 1,643 vs 1,643
- Exposure
- Comorbidity clusters in RA vs matched non-RA comparators
- RAn = 1,643
- Non-RAn = 1,643reference
- Outcome
- Clusters of comorbidities (44 morbidities and 11 chronic disease categories) identified by 4 unsupervised machine learning methods
Methodological detailspopulation-based, age-, sex-, and county-matched non-RA comparators, unsupervised machine learning (hierarchical clustering, factor analysis, K-means clustering, network analysis)
- Population-based
- Age-, sex-, and county-matched non-RA comparators
- Unsupervised machine learning (hierarchical clustering, factor analysis, K-means clustering, network analysis)
Main finding
As reported in the abstract
Comorbidity clusters
- Strong associations among mental/behavioral comorbidities and among cardiovascular risk factors and diseases
- Clusters associated with age and sex
- Differences between the 4 clustering methods driven by rare comorbidities and those weakly associated with other comorbidities; common comorbidities grouped together consistently across approaches
- Clusters unstable with different random seeds or bootstrap sampling
- Clusters of common comorbidities similar between RA and non-RA cohorts
Observational data: associations, not effects.
Original publication
Using Unsupervised Machine Learning Methods to Cluster Comorbidities in a Population‐Based Cohort of Patients With Rheumatoid Arthritis. Arthritis Care Res (Hoboken). Published 15 September 2022. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →